Source time Edition date
Close readLatent Space
[AINews] Kimi 3.1 Pro 104BA-2.8T: Fable-5.1-Class Open Weights narrows Closed-AI lead to 1 month... and Pro Highspeed API with 2-3× speed
[AINews] Kimi 3.1 Pro 104BA-2.8T：Fable-5.1级开放权重将封闭AI领先优势缩至1个月……以及2–3倍速的Pro Highspeed API
Kimi 3.1 Pro's open weights reportedly close the gap to closed models to about a month, so readers should weigh whether that pace of release makes open-weight deployment viable for their own stack.
Judgment
Platform and infra teams weighing open-weight standardization must now price a November 4 weight drop against unverified serving economics: Kimi K3.1 Pro's Highspeed tier is 2–3× faster but costs twice as much, so the cheaper token is not automatically the cheaper completed job.
Basis
Moonshot launched Kimi K3.1 Pro as a multimodal, "3T-class" mixture-of-experts model with 2.8 trillion total parameters, 104 billion active parameters and a 1-million-token context window, with open weights promised for November 4, 2026. Artificial Analysis placed it fourth on its Intelligence Index at 55, behind Opus 5.5 at 58 and Sonnet 5.5 and Muse Wonder Full at 56, with gains concentrated in agentic rather than knowledge and science tests. Arena ranked it #8 in Code Arena: WebDev at 1,692 points, up nine places and 37 points over Kimi K3. On Artificial Analysis' Cyber Index it scored 64%, the highest publicly accessible result, behind trusted-access GPT-6 Sol Daybreak Blue at 69%, despite deliberately reduced cyber training. Moonshot claims Highspeed preserves intelligence while generating output 2–3× faster, and its official account says it costs twice as much. The mechanism that decides viability is therefore served throughput per dollar on your own long-running agent workloads, not headline benchmark rank.
- Opus 5.558
- Sonnet 5.556
- Muse Wonder Full56
- Kimi K3.1 Pro55
Unclear The source gives no per-token price, latency, or throughput figures for Standard versus Highspeed, so no one can compute cost per completed task from it.
Next Check whether Moonshot actually publishes K3.1 Pro weights on November 4, 2026, and whether independent throughput tests confirm the 2–3× speedup at twice the price.