{"rewrite":{"id":"r_63ada337104585a678d98e47","clusterId":"c_f2265759eb57a94e257e26c7","slug":"thinking-machines-lab-releases-inkling-small-at-a-quarter-the-size","model":"deepseek-v4-flash:free","headline":"Thinking Machines Lab Releases Inkling-Small at a Quarter the Size","summary":"Thinking Machines Lab released Inkling-Small, an open-weight MoE model with 276 billion total and 12 billion active parameters. It delivers performance comparable to its 975-billion-parameter predecessor Inkling while using far fewer computational resources. The model is available on the Tinker service and as BF16 and NVFP4 weights on Hugging Face.","whyItMatters":"Inkling-Small shows that Thinking Machines Lab can compress its flagship model's performance into a fraction of the size, making high-end reasoning and audio capabilities more accessible.","webCardHtml":"\u003cp\u003eInkling-Small is a Mixture-of-Experts Transformer with 276 billion total parameters and 12 billion active, trained on NVIDIA GB300 NVL72. It keeps Inkling\u0026#39;s native reasoning for audio and images, variable thinking load, and a 1 million token context window.\u003c/p\u003e\u003cp\u003eBenchmark overlays show Inkling-Small and Inkling scoring similarly across ten tests, with one notable gap: Inkling-Small lags on the SimpleQA Verified factuality benchmark. Thinking Machines Lab says the smaller model performs on par or better on reasoning and agent tasks.\u003c/p\u003e\u003cp\u003eThe model inherits Inkling\u0026#39;s post-training safety measures, scoring competitively on the FORTRESS and StrongREJECT safety benchmarks. It is available on the Tinker service, with BF16 and NVFP4 weights on Hugging Face.\u003c/p\u003e","blueskyPost":"Thinking Machines Lab released Inkling-Small, an open-weight MoE model with 276B total params. It matches its 975B predecessor on most benchmarks, with a gap on factuality. Available on Tinker and Hugging Face.","twitterPost":"Thinking Machines Lab released Inkling-Small, an open-weight MoE with 276B total params. It matches its 975B predecessor on most benchmarks, with a gap on factuality. Available on Tinker and Hugging Face.","threadsPost":null,"newsletterBlurb":"Thinking Machines Lab released Inkling-Small, an open-weight MoE model with 276 billion total parameters. It delivers performance comparable to its 975-billion-parameter predecessor Inkling while using far fewer resources, with a notable gap on factuality benchmarks. The model is available on the Tinker service and on Hugging Face.","attributionJson":"[{\"source\":\"GIGAZINE\",\"url\":\"https://gigazine.net/news/20260731-thinking-machines-lab-inkling-small/\",\"title\":\"Thinking Machines Lab Releases 'Inkling-Small' Delivering Performance Comparable to Inkling at Nearly a Quarter the Size\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":4653,"outputTokens":595,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1786276703,"createdAt":"2026-08-09T11:53:46.000Z","publishedAt":"2026-08-09T11:56:44.000Z","updatedAt":"2026-08-09T11:53:46.000Z"},"cluster":{"id":"c_f2265759eb57a94e257e26c7","canonicalTitle":"Inklingと同等の性能をほぼ4分の1サイズで発揮する「Inkling-Small」をThinking Machines Labがリリース","representativeArticleId":"a_55897a7977f6d0ba24a043af","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[\"Inkling\",\"Inkling-Small\"],\"studios\":[\"Thinking Machines Lab\"],\"people\":[],\"type\":\"announcement\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-07-31T03:40:00.000Z","lastSeenAt":"2026-07-31T03:40:00.000Z","updatedAt":"2026-08-09T11:56:45.000Z"},"attribution":[{"source":"GIGAZINE","url":"https://gigazine.net/news/20260731-thinking-machines-lab-inkling-small/","title":"Inklingと同等の性能をほぼ4分の1サイズで発揮する「Inkling-Small」をThinking Machines Labがリリース"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":["Inkling","Inkling-Small"],"studios":["Thinking Machines Lab"],"people":[],"type":"announcement","domain":"other","is_roundup":false},"keyFacts":null}