https://vps.starsynced.net/Latest frontier model releases, refreshed every five minutes.Sat, 12 Sep 2026 18:11:54 +00005https://openai.com/index/perplexity-improving-accuracy-with-astrahttps://openai.com/index/perplexity-improving-accuracy-with-astraPerplexity uses Astra to write communications, change software, and monitor production systems, and checks in much less frequently than with earlier models.OpenAIMon, 14 Sep 2026 00:00:00 +0000https://pastagi.com/use-cases/devin-gpt-6-astra-self-testing-review-bottleneck/https://pastagi.com/use-cases/devin-gpt-6-astra-self-testing-review-bottleneck/Devin GPT-6 Astra self-testing shifts code review from reading diffs to auditing evidence. Here is the pattern, its failure modes, and audit criteria.PastagiSat, 12 Sep 2026 15:07:14 +0000https://towardsdatascience.com/one-capital-letter-was-silently-breaking-my-ai-support-bot-and-it-wasnt-in-the-new-model/https://towardsdatascience.com/one-capital-letter-was-silently-breaking-my-ai-support-bot-and-it-wasnt-in-the-new-model/A real Weave project that regression-tests three OpenAI models against the exact reply format your app depends on.TowardsdatascienceSat, 12 Sep 2026 14:00:02 +0000https://www.wired.com/story/security-news-this-week-from-hacks-to-bioweapons-claude-misuse-is-now-everywhere/https://www.wired.com/story/security-news-this-week-from-hacks-to-bioweapons-claude-misuse-is-now-everywhere/Plus: The US disrupts the internet’s biggest black market, a Conti ransomware hacker gets prison time, Meta fails to stop AI-generated videos of child abuse.WiredSat, 12 Sep 2026 10:30:00 +0000https://www.latent.space/p/ainews-deepseek-v41-flash-763b-p8bhttps://www.latent.space/p/ainews-deepseek-v41-flash-763b-p8bWe agree with Sebastian: this should have been DeepSeek v5LatentSat, 12 Sep 2026 05:56:05 +0000https://arxiv.org/abs/2609.10657https://arxiv.org/abs/2609.10657arXiv:2609.10657v1 Announce Type: new Abstract: Neural networks trained past memorization frequently undergo a delayed transition to generalization, a phenomenon known as grokking. Despite theoretical progress on \emph{why} this transition occurs, the quantitative structure of…ArxivSat, 12 Sep 2026 04:00:00 +0000https://arxiv.org/abs/2609.10712https://arxiv.org/abs/2609.10712arXiv:2609.10712v1 Announce Type: new Abstract: We study how model post-training and test-time inference design affect natural-language proof generation for hard olympiad mathematics. Starting from Nemotron 3 Ultra, we train two specialist checkpoints using supervised…ArxivSat, 12 Sep 2026 04:00:00 +0000https://www.marktechpost.com/2026/09/11/anthropic-adds-plugin-evals-to-claude-code-6-grader-types-a-no-plugin-baseline-and-a-ci-gate-for-skills/https://www.marktechpost.com/2026/09/11/anthropic-adds-plugin-evals-to-claude-code-6-grader-types-a-no-plugin-baseline-and-a-ci-gate-for-skills/Anthropic has published a new plugin evals workflow for Claude Code. The claude plugin eval command runs a plugin against realistic prompts, grades what Claude produced, and compares the result with a run where the plugin is not loaded. It answers 3 questions plugin developers…MarkTechPostFri, 11 Sep 2026 21:05:55 +0000https://techcrunch.com/2026/09/11/y-combinators-garry-tan-wants-u-s-open-weight-ai-labs-to-distill-frontier-models-too/https://techcrunch.com/2026/09/11/y-combinators-garry-tan-wants-u-s-open-weight-ai-labs-to-distill-frontier-models-too/Tan argues that frontier models themselves trained on public human knowledge so access to capable AI should be "a form of public good."TechcrunchFri, 11 Sep 2026 20:59:47 +0000https://techcrunch.com/2026/09/11/kimi-maker-moonshot-ai-targets-2-billion-in-annual-revenue/https://techcrunch.com/2026/09/11/kimi-maker-moonshot-ai-targets-2-billion-in-annual-revenue/While K3's usage figures have declined slightly in recent months, OpenRouter data currently shows as many as 300 billion tokens being generated each day by K3 models on the system.TechcrunchFri, 11 Sep 2026 19:35:54 +0000https://sakana.ai/fugu-max-release/https://sakana.ai/fugu-max-release/Sakana AI released Fugu Max and Fugu Ultra v2, two versions of its orchestrator that routes each task to other models. Fugu Max costs $2 and $6 per million tokens; Fugu Ultra v2 scores 48.3 on Chartography against Opus 5's 27.3.Sakana AIFri, 11 Sep 2026 12:00:00 +0000https://arxiv.org/abs/2609.11929https://arxiv.org/abs/2609.11929SenseNova-U1.5 is SenseTime's 8B-MoT unified multimodal model, and its technical report is now public. The paper covers an encoder-free, VAE-free design that reads and generates images at native resolutions up to 4K.SenseTimeThu, 10 Sep 2026 12:00:00 +0000https://developers.openai.com/api/docs/models/gpt-live-1https://developers.openai.com/api/docs/models/gpt-live-1GPT-Live-1 is now in the OpenAI API. The full-duplex voice model listens and speaks at once, handles interruptions, and hands deeper reasoning to the models and tools you pair it with. Voice sessions cost $0.05 per minute.OpenAIThu, 10 Sep 2026 12:00:00 +0000https://cohere.com/blog/north-small-translatehttps://cohere.com/blog/north-small-translateNorth Small Translate is a 218B-parameter open-weights translation model from Cohere Labs with 25B active parameters. It scores 83.60 on WMT26 across all languages, ahead of DeepL NextGen at 81.37.CohereThu, 10 Sep 2026 12:00:00 +0000https://cognition.com/blog/swe-2https://cognition.com/blog/swe-2SWE-2 is Cognition's new coding model, post-trained from Moonshot's 2.8-trillion-parameter Kimi K3. It scores 50.0% on FrontierCode 1.1 Main against 50.9% for Fable 5.1, and Cognition says it costs 64% less to run at that score.CognitionThu, 10 Sep 2026 12:00:00 +0000https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flashhttps://huggingface.co/deepseek-ai/DeepSeek-V4.1-FlashDeepSeek V4.1 Flash is now generally available under the API name deepseek-flash, with MIT open weights on Hugging Face. It scores 90.6 on Terminal-Bench 2.1 and 74.2 on DeepSWE v1.1, ahead of Opus 5.0 and GPT-5.6 Sol.DeepSeekThu, 10 Sep 2026 12:00:00 +0000https://huggingface.co/m-a-p/YuE2-3Bhttps://huggingface.co/m-a-p/YuE2-3BYuE2-3B turns lyrics and a style prompt into a full 48 kHz song with vocals. It writes an editable melody-and-chord score first, then renders the audio, and scores 6.96 on WildSongBench against Suno v6's 6.56.Multimodal Art ProjectionWed, 09 Sep 2026 12:00:00 +0000https://arxiv.org/abs/2609.10715https://arxiv.org/abs/2609.10715NCP-ArchPreview is an 8.9B open-weight language model from Shanghai AI Lab that predicts multi-token "concepts" alongside normal next-token prediction. It reaches OLMo-3-7B's final pretraining loss using 51.3% of the training tokens.Shanghai AI LabWed, 09 Sep 2026 12:00:00 +0000https://arxiv.org/abs/2609.10522https://arxiv.org/abs/2609.10522Show-Harness is an open control layer that exposes a robot as discrete semantic action units a vision-language model can reason over. Show Lab at NUS shipped the Apache-2.0 code, six LoRA adapters and the demonstration data with the paper.Show Lab, NUSWed, 09 Sep 2026 12:00:00 +0000https://suno.com/blog/introducing-v6https://suno.com/blog/introducing-v6Suno v6 is a new family of music models built with data licensed from Warner Music Group, BMG and Believe. It ships as v6, v6-wild and v6-mini, and Suno says it will retire its older models as the rollout finishes.SunoWed, 09 Sep 2026 12:00:00 +0000https://github.com/Omni-Interaction-Gander/Omni-Interaction-Agenthttps://github.com/Omni-Interaction-Gander/Omni-Interaction-AgentGander is an open 9B omni-interaction model that takes streaming video, speech and text together. You can interrupt it mid-sentence, and a separate reasoning agent keeps working on long tasks in the background.Tencent Hunyuan Speech TeamWed, 09 Sep 2026 12:00:00 +0000https://github.com/nex-agi/Nex-N2.5https://github.com/nex-agi/Nex-N2.5Nex-N2.5 is a family of three open-weight agent models from Nex AGI. The 1.6-trillion-parameter Max tier scores 92.6 on BrowseComp and 86.1 on Terminal-Bench 2.1. All three are Apache-2.0, and mini and Pro run free on OpenRouter.Nex AGITue, 08 Sep 2026 12:00:00 +0000https://github.com/Tencent-Hunyuan/AuKhttps://github.com/Tencent-Hunyuan/AuKAuK is an MIT-licensed 1.5B speech model from Tencent Hunyuan, Shanghai Jiao Tong University and the Shanghai Innovation Institute. It generates and edits speech from plain-language instructions. Code and weights shipped on September 8, 2026.Tencent HunyuanTue, 08 Sep 2026 12:00:00 +0000https://arxiv.org/abs/2609.08183https://arxiv.org/abs/2609.08183NeoHorse-1 is a pair of Apache-2.0 models, 4B and 9B, that TokenRhythm post-trained by routing agent work across a pool of models and turning the results into training data. The 4B scores 64.87 on the team's ten-benchmark average, up from 58.94.TokenRhythmTue, 08 Sep 2026 12:00:00 +0000