What shipped |
● Google · Gemini Pro · Oct 13 Google's first Gemini 4 model replaces Gemini 3.1 Pro as its top model in the paid Gemini API, with a 1M-token window shared by input, thinking and output. Compared with Gemini 3.1 Pro · the model it replaces 87% ✦ Google gives no earlier Gemini score to compare Answers nearly 9 in 10 questions about reasoning in the physical world · ERQA | $2 / $10 ✦ introductory price · Google's standard rate is $4 / $20 Launches at an introductory price for developers · API price |
|
● Meta · Muse Spark (open weights) · open weights · Oct 15 Meta's first downloadable Muse Spark: an 816B-parameter mixture-of-experts model with about 42B active parameters, a 1M-token context and text, image, video and audio input, released under the Muse Model License. Compared with GLM 5.3 · a competitor — comparison chosen by Meta 89.2 ▲ vs 84.2 for GLM 5.3 · +5 points · Meta's table, Muse Spark Open at xhigh effort Completes more tool-using agent tasks than GLM 5.3 · MCP-Atlas | 39 ✦ Xhigh effort · first open-weights Muse Spark, nothing earlier to compare Scores 39 on a combined set of hard tests · Artificial Analysis Intelligence Index v4.3.2 |
|
● Meta · Muse Wonder · Oct 12 Meta's first frontier-tier model, a larger step above Muse Spark built for agents, scoring 4 points higher on the Artificial Analysis index at about three to five times the API price. Compared with Muse Spark 1.3 · Meta's previous top model 52 ▲ up from 48 for Muse Spark 1.3 · +4 points · both at Max effort Scores higher than Muse Spark 1.3 on a combined set of hard tests · Artificial Analysis Intelligence Index v4.3.2 | $4 / $20 was $1.25 / $4.25 for Muse Spark 1.3 Costs developers several times more than Muse Spark 1.3 · API price |
|
Also new this week |
● DeepSeek · DeepSeek Flash · open weights · Oct 17 An update to DeepSeek-V4.1-Flash that scores higher on every agentic, coding and reasoning test DeepSeek published while giving shorter answers, with the same MIT weights, 1M-token context and API price. |
● Kuaishou · Kling · video model · Oct 15 |
● Black Forest Labs · FLUX [klein] · image model · open weights · Oct 15 Brings FLUX 3 Image's editing, up to ten reference images and bounding-box layout to a 5.6B open-weights model that runs on consumer GPUs down to an RTX 3090, with output up to 4K. |
● NVIDIA · Nemotron Super · open weights · Oct 15 |
● MiniMax · MiniMax M Flash · released Sep 27 · added this week |
Coming next |
Announced by the maker, not out yet |
“Grok 4.8, which is a 2.5T model trained with our new C++ software stack, will finish training this week and start RL” Elon Musk on X |
“all deepseek-v4-pro requests will route to V4.1-Flash at V4.1-Flash rates. This will continue until V4.1-Pro launches.” DeepSeek API docs |
● Qwen 4 · Alibaba · announced Sep 22 “At the conference, Alibaba revealed that its next-generation model, Qwen 4, is currently in training.” Alibaba Cloud |
● Qwen-Image 3.1 · Alibaba · announced Sep 22 · expected later this year “Qwen-Image 3.1, an image-generation model optimized for creative design and e-commerce marketing, is set to launch later this year.” Alibaba Cloud |
Close to their usual gap Time since the last release is close to the usual gap between this line’s models. We don’t predict dates. |
102 days since the last release · usual gap 114 days |
● Seed Lite · ByteDance · last: Doubao-Seed-2.0-Lite 247 days since the last release · usual gap 252 days |
140 days since the last release · usual gap 143 days |
173 days since the last release · usual gap 179 days, from only 2 gaps |
|
Next AI Model checks 17 AI makers for new models every morning. Every number comes from the maker or an independent test, with its source. One email a week, only when something ships. |
