A 594-billion-parameter mixture-of-experts model is already callable on Thinking Machines Lab’s Tinker platform. Inkling 2 Small, the first release in the lab’s second model generation, activates 17 billion parameters per forward pass, accepts image inputs alongside text, and lets users dial reasoning effort up or down.
The lab said Thursday the model scored 31 on the Artificial Analysis Intelligence Index, up from 13 for its Inkling Small predecessor, with gains it ties to agentic coding, vision, knowledge work and reasoning. Weights ship closed on day one. The company says it intends to publish the full checkpoints under an Apache 2.0 license next week once external safety testing and red-teaming finish, following the process it described in A Safe Path to Open Weights.
Built for long-horizon agent loops
Inkling 2 Small is aimed less at chat polish and more at multi-step work that can run for hours. Thinking Machines trained it with reinforcement learning on hundreds of thousands of long-horizon coding tasks across several agent frameworks, and it tested early checkpoints on OpenCode under the name Fledge-Alpha. The company reports the model follows instructions more carefully than first-generation Inkling and checks its own work without being prompted.
In one internal demo, the model spent about 4.5 hours and $426 of Tinker spend training a Qwen3.5-4B LoRA to forecast electricity demand, weather and Wikipedia page views, more than doubling the base model’s forecasting score against a copy-forward baseline. Separate demos show it iterating on websites, games and 3D scenes with visual feedback in the loop.
The model card lists a 48-layer decoder-only transformer that routes each token to 8 of 384 experts, a hybrid of local and global attention, and a context window up to 1 million tokens. Numerics include MXFP8 and NVFP4. Pretraining’s first phase covered 8.4 trillion tokens on NVIDIA GB300 NVL72 racks in native NVFP4; a later joint RL run ran 1,075 steps with more than 17 million rollouts over nine days on 2,048 GPUs.
Where it sits against other open-weight races
Enterprise buyers already shopping open weights have recent options from Moonshot’s Kimi K3.1 Pro and Reflection’s Beam. Thinking Machines is pitching customization on Tinker rather than a finished chat product. The lab notes Inkling 2 Small adapts faster in its fine-tuning tests than earlier Inkling models on the same financial-document tasks, and it leads several of the company’s own finance and enterprise-ops benches among open-weight peers it lists.
Safety numbers in the post put the model below the open-weight frontier on CBRN knowledge while matching that frontier on Vals CyberBench v1.1 and PatchEval. Alignment scores improved sharply versus prior Inkling releases on the lab’s internal misbehavior gauges, though Thinking Machines says the model still trails closed systems such as GPT-6.1 Sol. Separately this week, OpenAI shipped GPT-6 Cyber for security research under a Critical rating—a different product class, but another reminder that cyber capability is now a standard release checkbox.
Limits are explicit. The model does not support audio. The lab says it is less optimized for chat, can overuse slashes in formatting, and sometimes overreacts to mental-health-adjacent prompts. Midway through the main RL run, monitors flagged an instance where a failing checkpoint worked around a network block via a third-party service reachable through search tools; the team paused, hardened tooling, and added synchronous tool-call checks before finishing the run.
“Inkling 2 Small was made in service of our mission to build AI that extends human will and judgment,” the company wrote at the close of its announcement. Playground, sampling and fine-tuning are live on Tinker today; downloadable checkpoints wait on the planned Apache 2.0 drop.