October 24, 2026 Kimi K3.1 Pro: 55 on the Artificial Analysis Intelligence Index, Three Behind Claude Opus 5.5 Open Source Benchmarks Claude
October 18, 2026 DeepSeek-V4.1-Flash-1017: Terminal-Bench 4.0 climbs from 31.2 to 42.3, at 763 billion parameters under MIT DeepSeek Models Open Weights Benchmarks Coding
October 14, 2026 Gemini 4 Argon ships to Pro and Ultra at an introductory $2 per million tokens Google Gemini Claude Opus Pricing Benchmarks
October 14, 2026 Jack Clark moves full recursive self-improvement from 2028 to 2027 Anthropic Jack Clark RSI Claude Opus Benchmarks
October 7, 2026 Mistral Large 4 scores 82% on a cyber test where Claude Opus 5.5 sits near zero Mistral Open Weights Benchmarks Security Releases
October 7, 2026 OpenAI publishes 722 maths manuscripts from an internal model, three hours of compute per result OpenAI Mathematics Lean Research Benchmarks
October 6, 2026 Beam: Reflection's first open model has 501 billion parameters, and the weights only arrive later this month Reflection AI Open Weights Models Benchmarks Coding
October 5, 2026 GPT-6 Astra Downloaded StarSkirmish's Best Human Bot and Ran That Instead OpenAI Claude Opus 5.5 Agenten Benchmarks Sicherheit
October 1, 2026 Gemini 4 Argon Leads 12 of 18 Benchmarks, Claude Opus 5.5 Keeps Terminal-Bench 4.0 Google Gemini Claude Opus Benchmarks Cybersecurity
September 30, 2026 GPT-6.1 Sol costs exactly what Claude Sonnet 5.5 costs, and is measured against Opus 5.5 OpenAI Claude Benchmarks Pricing
September 22, 2026 MiMo-V2.6-Pro costs one twenty-third of Claude Opus 5 Open Weights Models Benchmarks
September 20, 2026 Vals keeps its tests secret — and the model makers pay for the privilege Benchmarks Evaluation Startups
September 4, 2026 GPT-6 Astra is here: OpenAI takes aim at Fable 5.1 with the same price and different strengths OpenAI GPT-6 Astra Fable 5.1 Benchmarks
July 26, 2026 Opus 5, Independently Judged: What Simon Willison and the Benchmarks Say Anthropic Opus 5 Benchmarks
July 17, 2026 Kimi K3: 2.8 Trillion Parameters — and Suddenly a Chinese Model Costs as Much as Sonnet Open Source Modelle Benchmarks
July 9, 2026 OpenAI Audits SWE-Bench Pro — and Finds Nearly a Third of Tasks Broken OpenAI Benchmarks Coding Research
June 22, 2026 GPT-5.6 Launch Week: OpenAI Needs to Deliver OpenAI GPT-5.6 Model-Release Benchmarks Competition
May 22, 2026 Cursor Composer 2.5: Matching Opus 4.7 at One-Tenth the Cost Cursor Composer AI Coding Benchmarks Developer
April 25, 2026 GPT-5.5 vs Claude Opus 4.7: The Benchmark Showdown in Detail GPT-5.5 Claude Opus 4.7 Benchmarks Comparison
April 16, 2026 Nature Study: AI Agents Fail at Complex Scientific Tasks AI Agents Research Stanford Nature Benchmarks
April 10, 2026 GLM-5.1: The Open-Source Model That Works Autonomously for 8 Hours Open Source GLM Agents Benchmarks
March 19, 2026 DeepMind Wants to Measure AGI — and Launches a Hackathon to Build the Tests Google DeepMind AGI Benchmarks Research Kaggle
