Model results
TextQuests
Choose a shared evaluation protocol to compare model scores. Versions, sources and configurations stay separate.
Atlas snapshot captured 10/19/2026, 10:20:28 AM UTC. This is an import timestamp, not the evaluation date. Coverage may be incomplete.
Evaluation groups
Open a group to see its exact protocol and scores. Different configurations remain separate.
No model rankings are available in this snapshot. Reported results remain available below.
Other reported results
These reports have no model ranking. Open a group for its evidence and the reason positions are withheld.
- Claude Fable 5dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=claud…
CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=claude-fable-5-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. Alias-level identity only where the publisher does not expose a dated provider endpoint. result date not supplied; captured 2026-10-19.
- Claude Fable 5.1dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=claud…
CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=claude-fable-5-1-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. Alias-level identity only where the publisher does not expose a dated provider endpoint. result date not supplied; captured 2026-10-19.
- Claude Haiku 4.5dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2025-10-15; identity reference=https://www.anthropic.com/news/claude-haiku-4-5. Exact Haiku 4.5 release and official announcement dated 2025-10-15. Thinking 32k remains configuration metadata. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=haiku-4-5-thinking-32k; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Claude Opus 4.5dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2025-11-24; identity reference=https://www.anthropic.com/news/claude-opus-4-5. Exact Opus 4.5 release and official announcement dated 2025-11-24. Thinking 32k remains configuration metadata. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=opus-4-5-thinking-32k; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Claude Opus 4.6dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-02-05; identity reference=https://www.anthropic.com/news/claude-opus-4-6. Exact named release; linked official announcement dated 2026-02-05. Adaptive 64k remains configuration metadata. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=opus-4-6-adaptive-64k; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Claude Opus 4.7dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-04-16; identity reference=https://www.anthropic.com/news/claude-opus-4-7. Exact named release; linked official announcement dated 2026-04-16. Adaptive 64k remains configuration metadata. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=opus-4-7-adaptive-64k; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Claude Opus 4.8dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-05-30; identity reference=https://www.anthropic.com/news/claude-opus-4-8. Exact named release and official announcement identify Opus 4.8. Dashboard date 2026-05-30 conflicts with official announcement 2026-05-28; neither is a measurement date. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=opus-4-8-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Claude Opus 5dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=claud…
CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=claude-opus-5-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. Alias-level identity only where the publisher does not expose a dated provider endpoint. result date not supplied; captured 2026-10-19.
- Claude Opus 5.5dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=claud…
CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=claude-opus-5-5-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. Alias-level identity only where the publisher does not expose a dated provider endpoint. result date not supplied; captured 2026-10-19.
- Claude Sonnet 4dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2025-05-22; identity reference=https://www.anthropic.com/news/claude-4. Exact Sonnet 4 release, matching date and official Claude 4 launch. Original config uses claude-sonnet-4-0 with 16000 thinking tokens; executed dashboard config remains unverified. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=sonnet-4-thinking-16k; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Claude Sonnet 4.5dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2025-09-29; identity reference=https://www.anthropic.com/news/claude-sonnet-4-5. Exact Sonnet 4.5 release and official announcement dated 2025-09-29; captured maintained config declares 32000 thinking tokens but does not prove dashboard execution. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=sonnet-4-5-thinking-32k; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Claude Sonnet 4.6dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-02-17; identity reference=https://www.anthropic.com/news/claude-sonnet-4-6. Exact named release; linked official announcement dated 2026-02-17. Adaptive 64k remains configuration metadata. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=sonnet-4-6-adaptive-64k; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- DeepSeek V3.2dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2025-12-01; identity reference=https://api-docs.deepseek.com/news/news251201. Exact V3.2 release, matching date and official V3.2 announcement distinguish stable release from V3.2 Exp. Thinking remains configuration metadata. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=deepseek-v3.2-thinking; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- DeepSeek V4 Pro 0423dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-04-24; identity reference=https://api-docs.deepseek.com/news/news260424. Dashboard date 2026-04-24 and linked official V4 Preview announcement identify the initial April Pro release rather than August Pro GA. Atlas labels that initial release 0423. The one-day discrepancy is retained; endpoint revision remains unknown. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=deepseek-v4-pro; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Gemini 2.5 Flashdashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2025-06-17; identity reference=https://blog.google/products-and-platforms/products/gemini/gemini-2-5-model-family-expands/. The official June 17 launch explicitly announces stable Gemini 2.5 Flash and Pro, distinguishing them from Flash-Lite preview. The stable Flash stem and dashboard releaseDate support this release; executed provider revision remains unknown. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gemini-2-5-flash-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Gemini 2.5 Prodashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2025-06-17; identity reference=https://blog.google/products-and-platforms/products/gemini/gemini-2-5-model-family-expands/. The official June 17 launch explicitly announces stable Gemini 2.5 Pro and Flash, distinguishing them from Flash-Lite preview. The stable Pro stem and maintained high-effort configuration support this release. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gemini-2-5-pro-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Gemini 3 Flash Previewdashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2025-12-17; identity reference=https://blog.google/products/gemini/gemini-3-flash. Preview suffix, exact named release, matching date and official Gemini 3 Flash announcement establish release identity. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gemini-3-flash-preview-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Gemini 3.1 Pro Previewdashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-02-19; identity reference=https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-1-pro/. Preview suffix, release date and linked announcement identify the standard preview, not Custom Tools. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gemini-3.1-pro-preview-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Gemini 3.5 Flashdashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-05-20; identity reference=https://deepmind.google/models/model-cards/gemini-3-5-flash/. Official Gemini 3.5 Flash card confirms the exact unique release and publication 2026-05-19, matching Atlas. Dashboard date 2026-05-20 conflicts and is retained; endpoint revision remains unknown. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gemini-3.5-flash; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Gemini 3.7 Flashdashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-08-01; identity reference=https://deepmind.google/models/model-cards/gemini-3-7-flash/. Official Gemini 3.7 Flash card confirms the exact unique release and publication 2026-08-13, matching Atlas. Dashboard date 2026-08-01 conflicts and is retained; endpoint revision remains unknown. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gemini-3.7-flash-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Gemini 3.8 Flashdashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-09-02; identity reference=https://deepmind.google/models/model-cards/gemini-3-8-flash/. Dashboard URL is empty. Official Gemini 3.8 Flash model card independently confirms name and 2026-09-02 publication, matching dashboard and Atlas. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gemini-3.8-flash-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Gemini 4 Argondashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release identity supported by publisher modelCardUrl https://deepmind.google/models/model-cards/gemi…
Release identity supported by publisher modelCardUrl https://deepmind.google/models/model-cards/gemini-4-argon/ and releaseDate 2026-10-13; dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gemini-4-argon-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. Alias-level identity only where the publisher does not expose a dated provider endpoint. result date not supplied; captured 2026-10-19.
- GLM 5.1dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-04-06; identity reference=https://huggingface.co/zai-org/GLM-5.1/raw/main/README.md. Exact GLM 5.1 name and dashboard releaseDate match Atlas. Official zai-org/GLM-5.1 model card confirms the exact release; dashboard blog returns an app shell. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=glm-5.1; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- GLM 5.2dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-06-16; identity reference=https://huggingface.co/zai-org/GLM-5.2/raw/main/README.md. Exact GLM 5.2 name and dashboard releaseDate match Atlas. Official zai-org/GLM-5.2 model card confirms the exact release; dashboard blog returns an app shell. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=glm-5.2; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- GLM 5.3dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-08-14; identity reference=https://huggingface.co/zai-org/GLM-5.3/raw/main/README.md. Official zai-org/GLM-5.3 model card confirms the exact named release and matches Atlas huggingFaceId. Dashboard 2026-08-14 conflicts with Atlas 20260816; endpoint revision remains unknown. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=glm-5.3; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- GPT-4o (2024-11-20)dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gpt-4…
CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gpt-4o-2024-11-20; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. Alias-level identity only where the publisher does not expose a dated provider endpoint. result date not supplied; captured 2026-10-19.
- GPT-5dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2025-08-07; identity reference=https://openai.com/index/introducing-gpt-5/. Exact GPT-5 release, matching dashboard date and official launch dated 2025-08-07; high configuration retained. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gpt-5-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- GPT-5 Minidashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2025-08-07; identity reference=https://openai.com/index/introducing-gpt-5-for-developers/. The official August 7 developer launch explicitly names gpt-5-mini, with high reasoning effort corroborated by the maintained publisher configuration. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gpt-5-mini-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- GPT-5 Nanodashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2025-08-07; identity reference=https://openai.com/index/introducing-gpt-5-for-developers/. The official August 7 developer launch explicitly names gpt-5-nano and supports reasoning effort. The dashboard high suffix remains publisher configuration metadata. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gpt-5-nano-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- GPT-5.1dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2025-11-12; identity reference=https://developers.openai.com/api/docs/models/gpt-5.1. Official GPT-5.1 API page explicitly lists gpt-5.1 and snapshot gpt-5.1-2025-11-13, supporting the Atlas release. Dashboard 2025-11-12 matches the ChatGPT announcement but differs from the API snapshot date. The benchmark endpoint remains unknown. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gpt-5.1-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- GPT-5.2dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2025-12-11; identity reference=https://openai.com/index/introducing-gpt-5-2/. Exact GPT-5.2 release, matching dashboard date and official announcement dated 2025-12-11. High configuration is neither Pro nor Chat. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gpt-5.2-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- GPT-5.4dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-03-03; identity reference=https://developers.openai.com/api/docs/models/gpt-5.4. Official GPT-5.4 API page explicitly lists gpt-5.4 and snapshot gpt-5.4-2026-03-05, supporting the Atlas release. Dashboard 2026-03-03 conflicts with that date and is retained as publisher metadata. The benchmark endpoint remains unknown. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gpt-5.4-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- GPT-5.4 Minidashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-03-17; identity reference=https://openai.com/index/introducing-gpt-5-4-mini-and-nano/. Exact Mini release, matching dashboard date and official mini/nano announcement dated 2026-03-17. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gpt-5.4-mini-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- GPT-5.4 Nanodashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-03-17; identity reference=https://openai.com/index/introducing-gpt-5-4-mini-and-nano/. Exact Nano release, matching dashboard date and official mini/nano announcement dated 2026-03-17. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gpt-5.4-nano-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- GPT-5.5dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-04-23; identity reference=https://openai.com/index/introducing-gpt-5-5/. Exact GPT-5.5 release, matching date and official launch dated 2026-04-23. High configuration is not Pro. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gpt-5.5-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- GPT-5.6 Lunadashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-07-09; identity reference=https://openai.com/index/gpt-5-6/. Exact Luna variant and matching dashboard date; official GPT-5.6 launch dated 2026-07-09 names Sol, Terra and Luna separately. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gpt-5.6-luna-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- GPT-5.6 Soldashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-07-09; identity reference=https://openai.com/index/gpt-5-6/. Exact Sol variant and matching dashboard date; official GPT-5.6 launch dated 2026-07-09 names Sol, Terra and Luna separately. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gpt-5.6-sol-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- GPT-5.6 Terradashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-07-09; identity reference=https://openai.com/index/gpt-5-6/. Exact Terra variant and matching dashboard date; official GPT-5.6 launch dated 2026-07-09 names Sol, Terra and Luna separately. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gpt-5.6-terra-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- GPT-6 Astradashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-09-03; identity reference=https://developers.openai.com/api/docs/models/gpt-6-astra. Dashboard URL is empty. Official OpenAI GPT-6 Astra API page independently confirms exact unique model name. Dashboard 2026-09-03 matches Atlas. Official page exposes alias gpt-6-astra, not a dated invocation. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gpt-6-astra-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- GPT-6.1 Soldashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release identity supported by publisher modelCardUrl https://openai.com/index/introducing-gpt-6-1-so…
Release identity supported by publisher modelCardUrl https://openai.com/index/introducing-gpt-6-1-sol/ and releaseDate 2026-09-29; dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=gpt-6.1-sol-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. Alias-level identity only where the publisher does not expose a dated provider endpoint. result date not supplied; captured 2026-10-19.
- Grok 4.3dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-04-17; identity reference=https://docs.x.ai/developers/models/grok-4.3. Exact unique Grok 4.3 name and official Grok 4.3 API page support release identity. Dashboard 2026-04-17 differs from Atlas 20260430; dated endpoint remains unknown. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=grok-4-3; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Grok 4.5dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-07-08; identity reference=https://x.ai/news/grok-4-5. The official announcement confirms the exact unique Grok 4.5 release but is dated 2026-07-16. Dashboard releaseDate and Atlas ID use 2026-07-08. This discrepancy is retained as release metadata; no dated endpoint or measurement date is established. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=grok-4.5; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Grok 4.6dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-08-12; identity reference=https://x.ai/news/grok-4-6. Official Grok 4.6 announcement dated 2026-08-12 matches dashboard and confirms exact unique named release. Atlas canonical ID carries 20260810 while route creation is 2026-08-12. This is release-level mapping with the date discrepancy retained. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=grok-4.6; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Kimi K2 Thinkingdashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2025-11-06; identity reference=https://huggingface.co/moonshotai/Kimi-K2-Thinking/raw/main/README.md. The official moonshotai/Kimi-K2-Thinking model card matches the exact Atlas huggingFaceId and links the dashboard blog. The maintained publisher config names kimi-k2-thinking. Dashboard releaseDate matches the Atlas release identity. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=kimi-k2-thinking; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Kimi K2.5dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-01-29; identity reference=https://www.kimi.com/blog/kimi-k2-5.html. Exact unique Kimi K2.5 release and linked official K2.5 blog support release identity. Dashboard date 2026-01-29 differs from Atlas 0127 identifier; no dated invocation is claimed. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=kimi-k2.5-thinking; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Kimi K2.6dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-04-20; identity reference=https://www.kimi.com/blog/kimi-k2-6. Exact K2.6 release, matching date and official K2.6 blog. 64k remains configuration metadata. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=kimi-k2.6-64k; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Kimi K3dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-07-16; identity reference=https://platform.kimi.ai/docs/guide/kimi-k3-quickstart. Exact unique Kimi K3 release and official K3 API guide support release identity. Dashboard date 2026-07-16 differs from Atlas 20260715 identifier; no dated invocation is claimed. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=kimi-k3-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Muse Spark 1.1dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-07-09; identity reference=https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/. Exact Muse Spark 1.1 release and matching dashboard date; official page explicitly introduces Muse Spark 1.1 on 2026-07-09. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=muse-spark-1.1-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- Muse Wonderdashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2026-10-12; identity reference=https://research.meta.ai/blog/muse-wonder. Exact Muse Wonder release, matching date and official announcement dated 2026-10-12. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=muse-wonder-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- o3dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2025-04-16; identity reference=https://openai.com/index/introducing-o3-and-o4-mini. Exact o3 release, matching dashboard date and official o3/o4-mini launch. Original publisher config explicitly uses openai/o3 with high reasoning effort. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=o3-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- o3 Mini Highdashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2025-01-31; identity reference=https://openai.com/index/openai-o3-mini/. Exact high configuration and launch date match the Atlas high variant. Official o3-mini launch distinguishes high effort; no dated invocation is claimed. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=o3-mini-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- o4 Mini Highdashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · Release-level identity supported by exact named release and publisher metadata; dashboard releaseDat…
Release-level identity supported by exact named release and publisher metadata; dashboard releaseDate=2025-04-16; identity reference=https://openai.com/index/introducing-o3-and-o4-mini/. Dashboard lacks modelCardUrl. Official o3/o4-mini launch dated 2025-04-16 and original publisher o4-mini high-effort config support the Atlas high variant. Dated provider endpoint revision is not reported. CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=o4-mini-high; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. result date not supplied; captured 2026-10-19.
- R1dashboard-unversioned
TextQuests · Center for AI Safety (reviewed publisher results)
Single-model result · Positions withheld
Measurement date not reported
Scope · CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=deeps…
CAIS dashboard TextQuests progress field across the 25-game benchmark; publisher configuration=deepseek-r1; clue setting, step budget, run count and executed harness revision not reported in dashboard result metadata; do not merge with the separate textquests.ai no-clues or with-clues cohorts; not completion rate or harm. Alias-level identity only where the publisher does not expose a dated provider endpoint. result date not supplied; captured 2026-10-19.