Alibaba's Qwen3.8 Max scores 56 on the Artificial Analysis Intelligence Index at
- Source
- x.com
- Date
Alibaba's Qwen3.8 Max scores 56 on the Artificial Analysis Intelligence Index at $1.14 per task, but open weights leader Kimi K3 remains 1 point ahead at 25% lower cost per task ($0.86) @Alibaba_Qwen has released Qwen3.8 Max, which Alibaba states is a 2.4T total parameter MoE activating 95B parameters per forward pass. Alibaba has announced it plans to release the weights next week, a shift in strategy as it has typically kept its Max class of models proprietary. Once released, Qwen3.8 Max would be ~6x larger than Alibaba's largest open weights release to date (Qwen3.5 397B) and the second largest open weights model behind Kimi K3 (2.8T) Note: we earlier published results showing Qwen3.8 Max scoring 53 on the Artificial Analysis Intelligence Index. Those runs were affected by intermittent issues on the endpoint we were evaluating, and we have re-run all evaluations on Alibaba's public API endpoint Key takeaways: ➤ Qwen3.8 Max scores 56 on the Artificial Analysis Intelligence Index, up 10 points from Qwen3.7 Max (46). It is in line with Claude Opus 4.8 (max, 56) and sits second among labs from China, ahead of GLM-5.2 (max, 51) but behind Kimi K3 (max, 57) ➤ Qwen3.8 Max scores…
Context
Alibaba's Qwen3.8 Max lands one point behind the open-weights leader on Artificial Analysis's headline index. The independent evaluator's August 6 thread scores it 56 on the Artificial Analysis Intelligence Index, up from 46 for Qwen3.7 Max and level with Claude Opus 4.8. Kimi K3, which Artificial Analysis calls the leader, scores 57 at $0.86 per task, against $1.14 for Qwen3.8 Max.
Kimi K3 is the natural comparison because Alibaba said it would publish Qwen3.8 Max's weights the following week, which Artificial Analysis called a shift in strategy: Alibaba had typically kept its Max models proprietary. Alibaba describes the model as a design with 2.4 trillion total parameters, of which about 95 billion are active on each forward pass. Released openly, it would be the second-largest open-weights model after Kimi K3 and about six times the size of Alibaba's largest previous open release. NVIDIA reported on August 12 that Alibaba had released the weights.
The index combines nine evaluations, and the gains were uneven. Terminal-Bench v2.1 rose 6 points and CritPt 7, while AA-Omniscience fell 10 points because the model attempted more questions it could not answer, with its rate rising from 23% to 40%. Artificial Analysis also flags the 42% τ³-Bench Banking score as out of line with the model's other results, since it places Qwen3.8 Max ahead of models that beat it elsewhere.
The headline number has already been revised once. Artificial Analysis first published 53, then reran every on Alibaba's public API after intermittent problems with the endpoint it had tested. The 56 comes from that rerun, and a one-point gap on a single aggregate index is a narrow margin.
- open weights — A model whose trained parameters are published for anyone to download and run — unlike API-only models you can access but never possess.
- mixture-of-experts — A model built from many specialist sub-networks where only a few activate per token, giving big-model capability at small-model running cost.
- hallucination — When a model states something false with full confidence — inventing facts, citations, or APIs that don't exist.
- eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
Checking sign-in…
Loading comments…