{"id":12626,"label":"Open models in perpetual catch-up","z_score_24h":4.804221525416279,"velocity_1h":0.0,"velocity_6h":0.0,"velocity_24h":2.0,"velocity_7d":5.0,"member_count":39,"source_diversity":2,"angles":[{"hook":"Everyone quotes a single benchmark score to declare open models are 'catching up,' but that composite hides which specific capabilities are closing fast and which remain firmly closed-model territory. The difference matters enormously for what you can actually ship today.","title":"The Open-Closed Gap Is Not One Number — It's a Capability Portfolio","novelty":"Most coverage treats the gap as scalar and celebrates each new open release as a milestone. Almost no one has mapped which capability sub-dimensions (coding, reasoning, multimodal, instruction following) are actually converging vs. structurally lagging — that decomposition is the real signal for builders choosing a stack.","sources_to_quote":["interconnects.ai — 'Reading today's open-closed performance gap' (Artificial Analysis Intelligence Index breakdown)","interconnects.ai — 'Open models in perpetual catch-up' (GLM 5 / Z.ai analysis)","Artificial Analysis Intelligence Index composite benchmark methodology"]},{"hook":"OpenAI just deprecated its fine-tuning APIs, and the same week Fireworks and Baseten hit decacorn valuations. That's not a coincidence — it's the market telling you where the AI value chain is actually consolidating.","title":"Fine-Tuning Is Dead. Long Live Inference Infrastructure.","novelty":"These two events — fine-tuning deprecation and inference infra unicorn/decacorn explosion — are being reported as separate news items. Nobody has yet explicitly connected them as two sides of the same architectural shift: prompting + inference optimization replacing model customization as the dominant engineering paradigm.","sources_to_quote":["latent.space — '[AINews] The End of Finetuning' (OpenAI fine-tuning API deprecation)","latent.space — '[AINews] New AI Infra decacorns: Fireworks, Baseten' (valuation data)","OpenRouter funding trajectory (mentioned in latent.space piece)","OpenAI deprecation announcement"]},{"hook":"Every major open-weights model generating serious buzz in the last 12 months — GLM 5, Qwen 3, DeepSeek — has come from a Chinese lab. This is not a streak; it's a structural shift in who is subsidizing the open-model ecosystem and why.","title":"Why Chinese Open-Weight Labs Are Quietly Winning the Open-Source AI Race","novelty":"The conversation stays surface-level ('another great open model!') and misses the strategic incentive asymmetry: Chinese labs release weights because export-controlled compute and geopolitical pressure make ecosystem influence more valuable than API lock-in. Western commentary keeps treating each release as a surprise instead of an inevitable pattern.","sources_to_quote":["interconnects.ai — 'Open models in perpetual catch-up' (Z.ai GLM 5, pattern across 12 months)","interconnects.ai — 'Gemma 4 and what makes an open model succeed' (contrasting Western lab releases like Llama 4 fiasco, Gemma 4)","Qwen 3 release notes / Alibaba Cloud","DeepSeek technical report"]}],"angles_generated_at":"2026-05-29T08:54:03.321727+00:00","first_seen_at":"2026-05-28T22:27:20.871776+00:00","last_seen_at":"2026-06-23T07:02:46.710814+00:00","members":[{"title":"Open models in perpetual catch-up","url":"https://www.interconnects.ai/p/open-models-in-perpetual-catch-up","source_type":"substack_post","author":"Nathan Lambert","source_published_at":"2026-02-17T17:27:36+00:00","similarity":1.0,"added_at":"2026-05-29T08:38:24.387198+00:00"},{"title":"[AINews] New AI Infra decacorns: Fireworks, Baseten (with OpenRouter on the way)","url":"https://www.latent.space/p/ainews-new-ai-infra-decacorns-fireworks","source_type":"substack_post","author":null,"source_published_at":"2026-05-27T03:33:53+00:00","similarity":0.8925463663445161,"added_at":"2026-05-29T08:38:25.612371+00:00"},{"title":"[AINews] Satya on Loopcraft: Building Frontier Ecosystems","url":"https://www.latent.space/p/ainews-satya-on-loopcraft-building","source_type":"substack_post","author":"Latent.Space","source_published_at":"2026-06-16T02:29:29+00:00","similarity":0.8854829539717058,"added_at":"2026-06-16T04:02:38.307287+00:00"},{"title":"[AINews] Open Models, Model Labs vs Agent Labs, and What's Untrainable — Sarah Guo","url":"https://www.latent.space/p/ainews-open-models-model-labs-vs","source_type":"substack_post","author":null,"source_published_at":"2026-06-11T03:14:26+00:00","similarity":0.8806958508887559,"added_at":"2026-06-11T05:02:38.260033+00:00"},{"title":"[AINews] GLM > GPT? GLM-5.2 passes vibe check; Z.ai forecasts Open Fable by December","url":"https://www.latent.space/p/ainews-glm-gpt-glm-52-passes-vibe","source_type":"substack_post","author":null,"source_published_at":"2026-06-19T05:53:54+00:00","similarity":0.8748943285386753,"added_at":"2026-06-19T07:02:38.331902+00:00"},{"title":"[AINews] Loopcraft: The Art of Stacking Loops","url":"https://www.latent.space/p/ainews-loopcraft-the-art-of-stacking","source_type":"substack_post","author":null,"source_published_at":"2026-06-12T05:34:09+00:00","similarity":0.872776537572534,"added_at":"2026-06-12T07:02:38.258765+00:00"},{"title":"Reading today's open-closed performance gap","url":"https://www.interconnects.ai/p/reading-todays-open-closed-performance","source_type":"substack_post","author":"Nathan Lambert","source_published_at":"2026-04-20T18:25:02+00:00","similarity":0.8621051755310156,"added_at":"2026-05-29T08:38:25.612371+00:00"},{"title":"[AINews] The End of Finetuning","url":"https://www.latent.space/p/ainews-the-end-of-finetuning","source_type":"substack_post","author":null,"source_published_at":"2026-05-13T02:47:22+00:00","similarity":0.8619275728079425,"added_at":"2026-05-29T08:38:25.612371+00:00"},{"title":"Gemma 4 and what makes an open model succeed","url":"https://www.interconnects.ai/p/gemma-4-and-what-makes-an-open-model","source_type":"substack_post","author":"Nathan Lambert","source_published_at":"2026-04-03T16:57:36+00:00","similarity":0.8610597462036937,"added_at":"2026-05-29T08:38:25.612371+00:00"},{"title":"[AINews] All Model Labs are now Agent Labs","url":"https://www.latent.space/p/ainews-all-model-labs-are-now-agent","source_type":"substack_post","author":null,"source_published_at":"2026-05-23T04:21:17+00:00","similarity":0.8564106213954961,"added_at":"2026-05-29T08:38:25.612371+00:00"},{"title":"[AINews] not much happened today","url":"https://www.latent.space/p/ainews-not-much-happened-today-6b8","source_type":"substack_post","author":null,"source_published_at":"2026-06-06T04:34:47+00:00","similarity":0.8545829650559337,"added_at":"2026-06-06T06:02:38.328805+00:00"},{"title":"GLM-5.2 is the step change for open agents","url":"https://www.interconnects.ai/p/glm-52-is-the-step-change-for-open","source_type":"substack_post","author":"Nathan Lambert","source_published_at":"2026-06-22T14:52:45+00:00","similarity":0.8490523962215288,"added_at":"2026-06-22T16:02:38.331378+00:00"},{"title":"[AINews] SpaceX is already a $28B/yr Neocloud","url":"https://www.latent.space/p/ainews-spacex-is-already-a-28byr","source_type":"substack_post","author":null,"source_published_at":"2026-06-23T06:19:49+00:00","similarity":0.8449935000814993,"added_at":"2026-06-23T08:02:38.245874+00:00"},{"title":"Latest open artifacts (#19): Qwen 3.5, GLM 5, MiniMax 2.5 — Chinese labs' latest push of the frontier","url":"https://www.interconnects.ai/p/latest-open-artifacts-19-qwen-35","source_type":"substack_post","author":"Florian Brand","source_published_at":"2026-03-03T16:30:59+00:00","similarity":0.8426943880183942,"added_at":"2026-05-29T08:38:25.612371+00:00"},{"title":"[AINews] not much happened today","url":"https://www.latent.space/p/ainews-not-much-happened-today-7a8","source_type":"substack_post","author":null,"source_published_at":"2026-06-05T06:44:49+00:00","similarity":0.8355276230109433,"added_at":"2026-06-05T08:02:38.329448+00:00"},{"title":"Open and closed models are on different exponentials","url":"https://www.interconnects.ai/p/open-and-closed-models-are-on-different","source_type":"substack_post","author":"Nathan Lambert","source_published_at":"2026-06-01T13:03:48+00:00","similarity":0.8330044267760451,"added_at":"2026-06-01T14:47:06.194685+00:00"},{"title":"What comes next with open models","url":"https://www.interconnects.ai/p/the-next-phase-of-open-models","source_type":"substack_post","author":"Nathan Lambert","source_published_at":"2026-03-16T13:00:51+00:00","similarity":0.8306465235998431,"added_at":"2026-05-29T08:38:25.612371+00:00"},{"title":"My bets on open models, mid-2026","url":"https://www.interconnects.ai/p/my-bets-on-open-models-mid-2026","source_type":"substack_post","author":"Nathan Lambert","source_published_at":"2026-04-15T18:20:00+00:00","similarity":0.8278908095636508,"added_at":"2026-05-29T08:38:25.612371+00:00"},{"title":"[AINews] GLM-5.2: the top Frontend Coding model in the world, IndexShare for Speculative Decoding","url":"https://www.latent.space/p/ainews-glm-52-the-top-frontend-coding","source_type":"substack_post","author":null,"source_published_at":"2026-06-17T05:37:40+00:00","similarity":0.8263036427432638,"added_at":"2026-06-17T07:02:38.338247+00:00"},{"title":"[AINews] Founders and Forward Deployed Engineers","url":"https://www.latent.space/p/ainews-founders-and-forward-deployed","source_type":"substack_post","author":null,"source_published_at":"2026-05-30T01:57:15+00:00","similarity":0.8258646305399402,"added_at":"2026-05-30T03:00:41.943348+00:00"},{"title":"[AINews] Cognition raises $1B in $26B Series D","url":"https://www.latent.space/p/ainews-cognition-raises-1b-in-26b","source_type":"substack_post","author":null,"source_published_at":"2026-05-28T07:26:09+00:00","similarity":0.8237376823149477,"added_at":"2026-05-29T08:38:25.612371+00:00"},{"title":"[AINews] Reve 2 and Ideogram 4: Layouts in Imagegen","url":"https://www.latent.space/p/ainews-reve-2-and-ideogram-4-layouts","source_type":"substack_post","author":null,"source_published_at":"2026-06-04T03:24:07+00:00","similarity":0.8188646383232375,"added_at":"2026-06-04T05:02:38.320141+00:00"},{"title":"How open model ecosystems compound","url":"https://www.interconnects.ai/p/how-open-model-ecosystems-compound","source_type":"substack_post","author":"Nathan Lambert","source_published_at":"2026-05-12T15:54:47+00:00","similarity":0.8140416100707536,"added_at":"2026-05-29T08:38:25.612371+00:00"},{"title":"[AINews] NVIDIA Cosmos 3, Nemotron 3 Ultra, and RTX Spark","url":"https://www.latent.space/p/ainews-nvidia-cosmos-3-nemotron-3","source_type":"substack_post","author":null,"source_published_at":"2026-06-02T03:28:10+00:00","similarity":0.8120110382211627,"added_at":"2026-06-02T04:47:06.126389+00:00"},{"title":"[AINews] OpenAI GPT-next disproves 80 year old Erdős planar unit distance problem for under $1000","url":"https://www.latent.space/p/ainews-openai-gpt-next-disproves","source_type":"substack_post","author":null,"source_published_at":"2026-05-21T07:28:36+00:00","similarity":0.8107241039182939,"added_at":"2026-05-29T08:38:25.612371+00:00"},{"title":"[AINews] Fable and Mythos officially too dangerous to release","url":"https://www.latent.space/p/ainews-fable-and-mythos-officially","source_type":"substack_post","author":null,"source_published_at":"2026-06-13T04:30:52+00:00","similarity":0.8096961406983628,"added_at":"2026-06-13T06:02:38.338087+00:00"},{"title":"Latest open artifacts (#21): Open model bonanza! Gemma 4, DeepSeek V4, Kimi K2.6, MiMo 2.5, GLM-5.1 & others. On CAISI's V4 assessment.","url":"https://www.interconnects.ai/p/latest-open-artifacts-21-open-model","source_type":"substack_post","author":"Florian Brand","source_published_at":"2026-05-16T17:00:11+00:00","similarity":0.8088567011899112,"added_at":"2026-05-29T08:38:25.612371+00:00"},{"title":"[AINews] FrontierCode: Benchmarking for Code Quality over Slop","url":"https://www.latent.space/p/ainews-frontiercode-benchmarking","source_type":"substack_post","author":null,"source_published_at":"2026-06-09T06:12:33+00:00","similarity":0.8065536116497195,"added_at":"2026-06-09T08:02:38.274008+00:00"},{"title":"Claude Fable 5 and new AI safety fables","url":"https://www.interconnects.ai/p/claude-fable-5-and-new-ai-safety","source_type":"substack_post","author":"Nathan Lambert","source_published_at":"2026-06-09T22:59:35+00:00","similarity":0.8060379817667614,"added_at":"2026-06-10T00:02:38.301353+00:00"},{"title":"Latest open artifacts (#20): New orgs! New types of models! With Nemotron Super, Sarvam, Cohere Transcribe, & others","url":"https://www.interconnects.ai/p/latest-open-artifacts-20-new-orgs","source_type":"substack_post","author":"Florian Brand","source_published_at":"2026-03-30T13:02:45+00:00","similarity":0.801406852720076,"added_at":"2026-05-29T08:38:25.612371+00:00"},{"title":"[AINews] Anthropic growing 10x/year while everyone else is laying off >10% of their workforce","url":"https://www.latent.space/p/ainews-anthropic-growing-10xyear","source_type":"substack_post","author":null,"source_published_at":"2026-05-09T01:08:28+00:00","similarity":0.7937376994254411,"added_at":"2026-05-29T08:38:25.612371+00:00"},{"title":"Some ideas for what comes next, May 2026","url":"https://www.interconnects.ai/p/some-ideas-for-what-comes-next-may","source_type":"substack_post","author":"Nathan Lambert","source_published_at":"2026-05-26T15:39:02+00:00","similarity":0.7877785100209145,"added_at":"2026-05-29T08:38:25.612371+00:00"},{"title":"[AINews] Codex Rises, Claude Meters Programmatic Usage","url":"https://www.latent.space/p/ainews-codex-rises-claude-meters","source_type":"substack_post","author":null,"source_published_at":"2026-05-14T03:53:26+00:00","similarity":0.7803421578120412,"added_at":"2026-05-29T08:38:25.612371+00:00"},{"title":"[AINews] Everything is Conductor","url":"https://www.latent.space/p/ainews-everything-is-conductor","source_type":"substack_post","author":null,"source_published_at":"2026-05-15T00:30:21+00:00","similarity":0.7761406132793431,"added_at":"2026-05-29T08:38:25.612371+00:00"},{"title":"Notes from inside China's AI labs","url":"https://www.interconnects.ai/p/notes-from-inside-chinas-ai-labs","source_type":"substack_post","author":"Nathan Lambert","source_published_at":"2026-05-07T15:42:43+00:00","similarity":0.7721938404332512,"added_at":"2026-05-29T08:38:25.612371+00:00"},{"title":"[AINews] How to land a job at a frontier lab (on Pretraining)","url":"https://www.latent.space/p/ainews-how-to-land-a-job-at-a-frontier","source_type":"substack_post","author":null,"source_published_at":"2026-05-19T07:31:40+00:00","similarity":0.7716842759051656,"added_at":"2026-05-29T08:38:25.612371+00:00"},{"title":"Can open-source beat OpenAI?","url":"https://restofworld.org/2026/tiezhen-wang-china-us-open-source-ai/?utm_source=rss&utm_medium=rss&utm_campaign=feeds-full","source_type":"tech_article","author":"Kinling Lo","source_published_at":"2026-06-15T10:00:00+00:00","similarity":0.7638190312150207,"added_at":"2026-06-15T12:02:38.357832+00:00"},{"title":"The inevitable need for an open model consortium","url":"https://www.interconnects.ai/p/the-inevitable-need-for-an-open-model","source_type":"substack_post","author":"Nathan Lambert","source_published_at":"2026-04-11T13:02:06+00:00","similarity":0.7622750834447691,"added_at":"2026-05-29T08:38:25.612371+00:00"},{"title":"The Professor of Outputmaxxing — Anjney Midha, AMP","url":"https://www.latent.space/p/anj","source_type":"substack_post","author":null,"source_published_at":"2026-06-18T17:30:00+00:00","similarity":0.762209299373451,"added_at":"2026-06-18T19:02:38.331504+00:00"}]}