Inkling 2 Small: Thinking Machines' 594B MoE, second-gen model with Apache 2.0 weights coming
Thinking Machines Lab released Inkling 2 Small, a 594B-total / 17B-active MoE with 1M-token context, scoring 31 on the Artificial Analysis Intelligence Index and 86.3% on SWE-Bench Pro v2 (Hard); closed weights on Tinker now, Apache 2.0 release planned next week.
Why it matters
A 594B model that nearly matches Claude Opus 5.5 on coding benchmarks while promising full open weights next week; its single joint RL run (17M+ rollouts, 9 days, 2,048 GPUs) also autonomously fine-tuned another model in 4.5 hours.
What you could build with it
An indie developer building a long-horizon coding agent could route planning and self-review through Inkling 2 Small on Tinker, using its 1M-token context to hold an entire codebase in one session while the model checks its own work before committing changes.
Does it hold up?
Too early to judge for builders: weights are closed until next week's Apache 2.0 release, so today it runs only on Tinker, where the lab's own demo autonomously fine-tuned a 4B model in 4.5 hours, but no independent team has shipped with it yet.
Built with Inkling 2 Small
- Thinking Machines Releases Inkling 2 Small, a 594B MoE With 17B Activemarktechpost · Technical breakdown of the 594B/17B-active MoE, 1M context, 86.3% SWE-Bench Pro v2 (Hard) and the planned Apache 2.0 weights release.
- Thinking Machines Lab Launches Inkling 2 Small with Open Weights Pendingunite.ai · Launch coverage of the 1,075-step joint RL run, the alignment controls that gate the open-weights release, and the autonomous Qwen3.5-4B fine-tuning demo.
- Inkling 2 Small: 594B parameters, 31 on the AI indexsiliconnoon · Analysis noting the 68-hour pretraining run on GB300s, the $0.30/$1.00 per 1M pricing, and the caveat that weights are not out yet and Tinker serves 64K tokens by default.
Learn more
First spotted on article: source.
More AI models
VoiceStudio: open-source, fully-local ElevenLabs alternative for voice cloning, dubbing, and audiobooksdebpalash/VoiceStudio is a desktop app that runs voice cloning, voice design, video dubbing, dictation, transcription…models · JEV 0.75Soup: fine-tune an 8B LLM on a 4 GB laptop GPU from one YAML fileAn open-source CLI that fine-tunes LLMs from a single YAML file, with layer streaming that trains an 8B model on a 4…models · JEV 0.73Tencent Hunyuan Hy3: open 295B MoE with 21B active under Apache 2.0Tencent open-sourced Hy3, a 295B-parameter mixture-of-experts model with only 21B parameters active per token, under…models · JEV 0.73NVIDIA Isaac Lab 3.0 GA + Isaac OS open runtime + Agile ONE SN humanoidMajor Isaac platform expansion at GTC Berlin Oct 21: Isaac Lab 3.0 is generally available as the open-source robot…models · JEV 0.73NVIDIA Nemotron 3.5 Retriever collection: Embed 8B/1B, Rerank 8B, Parse 631MNVIDIA's Oct 21 collection drop for agentic retrieval: Nemotron-3.5-Embed in 8B BF16, 1B BF16 and 1B NVFP4…models · JEV 0.72Kimi K3.1 Pro: Moonshot's 2.8T open-weight flagship officially launchesMoonshot AI officially launched Kimi K3.1 Pro on Oct 23, a mixture-of-experts model with 2.8 trillion total / 104…models · JEV 0.72
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.