Mira Murati's Inkling Small Matches Flagship at 1/3 Parameters
ArtificialAnlys · x · 2026-07-31
Thinking Machines, founded by former OpenAI CTO Mira Murati, has released Inkling Small (276B total parameters, 12B active, Apache 2.0). Despite using less than a third of the flagship's parameters, it scores 40 on the Artificial Analysis Intelligence Index—just one point behind its larger sibling—making it the highest-scoring open weights model at its size.
Key Evaluation Findings:
- Coding & Reasoning: Meets or exceeds the flagship on several benchmarks, including HLE (32% vs. 30%), GPQA Diamond (89% vs. 87%), and SciCode (49% vs. 46%).
- Agentic & Factual: Compromises on factual accuracy with an AA-Omniscience score of -9 (vs. 2), but compensates with better presentation in long-horizon agentic tasks, achieving higher Elo.
- Token Efficiency: Averaged 24K output tokens per task, slightly fewer than the flagship and significantly more efficient than peers like DeepSeek V4 Flash (45K) and GPT-5.4 mini (78K).
- Pricing & Speed: Priced at $0.30/1M input and $1.20/1M output tokens, with a solid generation speed of 92 tokens/s.
Related event: Inkling-Small: 276B params, 12B active, open-source tops ARC(48 posts)→
More from Models
- Reflection's 501B MoE Beam hits 463 tok/s on 8x GB200 with NVFP4 at day 0 — zhyncs42 · 2026-10-16
- JevBench v1.7.46 leaderboard: new d1-ultra tops 32 API systems at 76.1 composite — airesearch12 · 2026-10-16
- LMArena launches Agent Arena and Code Arena leaderboards for model comparison — arena · 2026-10-16
- Third-party tests find NVIDIA Nemotron-3.5 Bolt underperforms nano on pandas/numpy and decision-making tasks — norenEnmotalen · 2026-10-16
- Reflection open-sources Beam under Apache 2.0 with 40+ partner ecosystem — bhutanisanyam1 · 2026-10-16
- One Prompt, 629 Lines: IQuest-Q1 Builds a Deep-Sea Submarine Game in 2 Minutes — Div_pradeep · 2026-10-16