Vibeleaderboard
← Vibers / Companies
Arcee AI

Arcee AI

Notable people

People map

Current people

Notable public-facing people we cover. Grouped by public roles, not reporting lines; this is not a complete staff directory.

Leadership

Other public roles

Former people

Role status unverified

Writing, research & talks

  • Arcee is partnering with Bolt to train and evaluate future open-weight models on real developer sessions, giving opted-in Bolt Pro users 50x usage on models like GLM-5.3 and DeepSeek-v4-Pro for a month in exchange for using their building sessions as training data.

  • Arcee's open models API beta widens past Trinity to serve DeepSeek-V4 Pro and Flash, GLM-5.2, Kimi-K3 and Thinking Machines' Inkling-Small, with per-million-token pricing listed for each and shipped alongside its open-source nac agent harness.

  • Nacarcee.ai

    Arcee open-sources nac, an agent harness for tasks that run for hours or days. The design splits ephemeral action context from durable workstream state: an orchestrator that can only dispatch threads, which report back structured episodes instead of raw transcripts.

  • Loka, AWS and Arcee post-trained Trinity Mini across two reinforcement-learning environments for biomedical research: one rewarding tool choice and evidence-seeking, one rewarding Gene Ontology inference in strict JSON. 21 controlled runs, with prompt-search gains reported apart from held-out result

  • Trinity-Large-Thinking ships under Apache 2.0 with open weights, targeting the failure modes that break long agent runs: context coherence across turns, tool calling that stays disciplined, and instruction following under constraint. Its instruct predecessor stays free on OpenRouter.

  • Arcee and the Department of Energy will build Genesis-Science-1, a trillion-parameter-class open-weight model on the next Trinity base, paired with a governed execution system for long scientific tasks. Weights and a technical report are promised later this year.

  • Every Trinity model, including previously shipped weights and quantizations, moves to OpenMDW-1.1, a Linux Foundation license that covers weights, configs, evaluation materials, and data as a single distribution instead of stretching a code license over them. NVIDIA plans to follow.

  • Arcee opens a Model-Tools repository of LLMOps utilities: a GUI over the LM Eval Harness, a model difference examiner that reports shape and config mismatches between checkpoints, and an interactive base-model-plus-adapter merger.

  • A training report on distilling a 405B teacher into 70B and 8B students: why live distillation was infeasible, how compressing extracted logits cut the dataset from petabytes to 50GB, the FSDP training configuration used, and how a parallel SNR-targeted run was merged in.

  • Arcee's Orthogonal Matching Pursuit method transplants one model's tokenizer into another with no retraining, reconstructing unseen token embeddings from shared ones. Reported: 96% performance retention across families and vocabularies that overlap only 54%.

  • Adapting a base model to a specialist domain usually costs you its general ability. This case study pairs continual pre-training with model merging to keep both, applied to medical and patent corpora, with citations to prior domain-adapted models and the cost argument behind the approach.

  • AFM-4.5B is Arcee's first from-scratch foundation model, trained on 6.58 trillion tokens curated with DatologyAI. The post argues the 3–10B band had gone stale and details why data curation, not architecture, was the hard part.

  • DistillKit's technical paper covers two distillation routes: KL divergence over teacher logits, and intermediate hidden-state alignment. It details the constraints of each and the SFT-based experiments Arcee ran to compare them against vanilla training.

  • MergeKit gains evolutionary merging, an open implementation of the method Sakana published but never released. The tutorial covers GPU requirements, defining custom lm-eval scoring tasks instead of optimising against leaderboards, and running the search.

  • An unflattering, blow-by-blow account of stretching AFM-4.5B from 4k to 64k context: which long-context papers survived contact with reality, why RULER alone misleads, and how merging and soups recovered short-context performance.

  • A working MergeKit walkthrough using Llama-3 8B base and instruct as the subjects: authoring the merge YAML, running the merge on a GPU box, generating a model card, and pushing the result, with the GUI path as an alternative.

  • Arcee-Nova is a 72B open model built by merging Qwen2-72B-Instruct with a custom-tuned generalist and layering RLHF on top, released with GGUF quantisations and scored on the OpenLLM Leaderboard 2.0 evaluation stack.

  • Arcee Agent is a 7B open model built on Qwen2 for one job: interpreting and chaining tool calls. It speaks several tool-use formats and can sit as a router in front of bigger models, with the tradeoff of weaker general capability.

  • Arcee-VyLinh is a 3B Vietnamese model built from Qwen2.5-3B via EvolKit-generated hard questions answered by a 72B teacher, SFT, a merge with Qwen2.5-3B-Instruct, and DPO refinement, shipped with a quantised build for consumer hardware.

  • Arcee ships SuperNova-Medius, a 14B open-weight model distilled from Llama-3.1-405B across architectures, and MergeKit's author uses the release to lay out where synthetic-data, logit, and hidden-state distillation each break down.

  • A postmortem of Meta's Llama 4 launch: the Maverick checkpoint submitted to LM Arena differed from the public weights, Scout's advertised 10M-token context crashed and diverged between GPU and CPU runs, and denials landed alongside an admitted chat-optimized submission.

  • MergeKit v0.1 drops the architecture allowlist. It merges any transformers-supported model including vision-language ones, adds a mergekit-pytorch path for raw PyTorch checkpoints such as diffusion and audio models, and makes the Arcee Fusion selective merge method public.

  • Arcee benchmarks its AFM-4.5B model on Intel Sapphire Rapids, AWS Graviton4 and Qualcomm Z1E processors at several quantizations, and explains what changed to make CPU serving viable: accelerated instruction sets, mature quantization and serving stacks, and smaller models.

  • Arcee-SuperNova is a 70B model assembled from three separately trained variants: a distillation of Llama-3.1-405B-Instruct into a 70B, an instruction-tuned model built on synthetic data from the EvolKit pipeline, and a DPO-aligned version given less weight in the final merge.

Arcee AI | VibeLeaderboard