good to see you.
Daily briefing
Good morning. Benchmark day today - Gemini 3.6 Flash is putting up solid numbers, scoring 70.2 on knowledge, 73.7 on multimodal grounded, 64.02 on coding, and 14 on gdp-pdf, while also posting a strong arena score on lmarena. Its smaller sibling Gemini 3.5 Flash-Lite trails behind with 62.1 on knowledge and 62.9 on multimodal grounded. Meanwhile Kimi K3 quietly flexed a big 87.9 on multimodal grounded, and Amazon Nova 2 Pro logged a 2 on gdp-pdf. Enjoy your day!
Today
OpenAI outlines its account of GPT-5. 6 efficiency across inference and agentic workflows, with a focus on useful work per dollar rather than a single benchmark result.
OpenAI · Jul 29, 2026
Read at openai.comOpenAI reports that retaining reasoning and enabling compaction in its Responses API harness raised GPT-5. 6 Sol's ARC-AGI-3 public-set score from 13.
OpenAI · Jul 29, 2026
Read at openai.com Read OutYet reportGoogle says the model can plan multi-step tasks, orchestrate lower-level robotics tools, and use continuous video to track progress and adapt to failures.
Google DeepMind · Jul 30, 2026
Read at deepmind.googleAWS documents explicit prompt caching for GPT-5. 6 on Bedrock, including cache boundaries, a 30-minute reuse window, and discounted cache reads.
AWS · Jul 30, 2026
Read at aws.amazon.com Read OutYet reportOpenAI says GPT-5. 6 Sol's API Fast mode replaces Priority Processing, offering up to 2.
OpenAI · Jul 30, 2026
Read at openai.com Read OutYet reportReleases
Is Mistral Code Agent Latest out yet?
Yes
Released Jul 31, 2026
Mistral Code Agent Latest is a language model from Mistral AI, now publicly released. It is positioned as the latest version of the company's code agent offering.
Is Claude Opus 5 out yet?
Yes
Released Jul 24, 2026
Claude Opus 5 is Anthropic's latest language model, now publicly released. It is the newest entry in the company's Opus line of Claude models.
Is GPT-5.6 Luna out yet?
Yes
Released Jul 9, 2026
OpenAI's GPT-5.6 Luna model, tracked with official availability evidence and vendor-reported HealthBench results.
Is GPT-5.6 Sol out yet?
Yes
Released Jul 9, 2026
OpenAI's next-generation model with stronger capabilities in coding, science, and cybersecurity, paired with an advanced safety stack.
Is GPT-5.6 Terra out yet?
Yes
Released Jul 9, 2026
OpenAI's GPT-5.6 Terra model, tracked with official availability evidence and vendor-reported HealthBench results.
Is GPT-Live out yet?
Yes
Released Jul 9, 2026
A new generation of voice models for natural human-AI interaction now powering ChatGPT Voice.
Is Grok 4.5 out yet?
Yes
Released Jul 8, 2026
Grok 4.5 is an AI model from xAI, now publicly released. Prior to release, it entered private beta testing at Tesla and SpaceX; detailed information about its capabilities has not been provided in official announcements.
Watchlist
Follow labs or models from the board and they'll land here.
ManageBenchmarks
Highest current score
Anthropic
Captured 2026-08-06. LM Arena source
Second highest
Anthropic
Captured 2026-08-06. LM Arena source
