AI model changelog

GPT-5.5 — scored 78.8 on RunFree Score

Claude Opus 4.8 — scored 94.7 on RunFree Score

Claude Opus 4.7 — scored 85.7 on RunFree Score

Claude Sonnet 4.6 — scored 66.2 on RunFree Score

Gemini 3.1 Pro Preview — scored 78 on RunFree Score

Gemini 3 Flash Preview — scored 63.5 on RunFree Score

Gemini 2.5 Pro — scored 22.3 on RunFree Score

GLM 5.2 — scored 86.3 on RunFree Score

DeepSeek V4 Pro — scored 72.4 on RunFree Score

DeepSeek V4 Flash — scored 59.1 on RunFree Score

Kimi K2 Thinking — scored 45.2 on RunFree Score

Llama 4 Maverick — scored 0 on RunFree Score

Qwen3 Max Thinking — scored 94.8 on RunFree Score

Fugu Ultra — Sakana added to the leaderboard

North Mini Code (free) — Cohere added to the leaderboard

GLM 5.2 — Z.ai added to the leaderboard

GLM 5.2 — scored 91.2% on GPQA Diamond

GLM 5.2 — scored 40.5% on Humanity's Last Exam

GLM 5.2 — scored 81% on Terminal-Bench

Kimi K2.7 Code — MoonshotAI added to the leaderboard

Claude Fable Latest — Anthropic added to the leaderboard

Claude Fable 5 — Anthropic added to the leaderboard

Nex-N2-Pro — Nex AGI added to the leaderboard

Nemotron 3.5 Content Safety (free) — NVIDIA added to the leaderboard

Nemotron 3 Ultra (free) — NVIDIA added to the leaderboard

Nemotron 3 Ultra — NVIDIA added to the leaderboard

Qwen3.7 Plus — Qwen added to the leaderboard

MiniMax M2.7 — scored 57% on Terminal-Bench

MiniMax M3 — MiniMax added to the leaderboard

Step 3.7 Flash — StepFun added to the leaderboard

GPT-5.5 — scored 78.2% on Terminal-Bench

GPT-5.5 — scored 41.4% on Humanity's Last Exam

GPT-5.5 Pro — scored 43.1% on Humanity's Last Exam

Claude Opus 4.8 — scored 88.6% on SWE-Bench Verified

Claude Opus 4.8 — scored 74.6% on Terminal-Bench

Claude Opus 4.8 — scored 93.6% on GPQA Diamond

Claude Opus 4.8 — scored 49.8% on Humanity's Last Exam

Claude Opus 4.7 — scored 87.6% on SWE-Bench Verified

Claude Opus 4.7 — scored 66.1% on Terminal-Bench

Claude Opus 4.7 — scored 94.2% on GPQA Diamond

Claude Opus 4.7 — scored 46.9% on Humanity's Last Exam

Gemini 3.1 Pro Preview — scored 44.4% on Humanity's Last Exam

Claude Opus 4.8 (Fast) — Anthropic added to the leaderboard

Claude Opus 4.8 — Anthropic added to the leaderboard

Qwen3.7 Max — Qwen added to the leaderboard

Grok Build 0.1 — xAI added to the leaderboard

Gemini 3.5 Flash — Google added to the leaderboard

Claude Opus 4.7 (Fast) — Anthropic added to the leaderboard

Perceptron Mk1 — Perceptron added to the leaderboard

Ring-2.6-1T — inclusionAI added to the leaderboard

Gemini 3.1 Flash Lite — Google added to the leaderboard

GPT Chat Latest — OpenAI added to the leaderboard

Grok 4.3 — xAI added to the leaderboard

Granite 4.1 8B — IBM added to the leaderboard

Mistral Medium 3.5 — Mistral added to the leaderboard

Nemotron 3 Nano Omni (free) — NVIDIA added to the leaderboard

Laguna XS.2 (free) — Poolside added to the leaderboard

Laguna XS.2 — Poolside added to the leaderboard

Laguna M.1 (free) — Poolside added to the leaderboard

Laguna M.1 — Poolside added to the leaderboard

Anthropic Claude Haiku Latest — Anthropic added to the leaderboard

OpenAI GPT Mini Latest — OpenAI added to the leaderboard

Google Gemini Pro Latest — Google added to the leaderboard

MoonshotAI Kimi Latest — MoonshotAI added to the leaderboard

Google Gemini Flash Latest — Google added to the leaderboard

Anthropic Claude Sonnet Latest — Anthropic added to the leaderboard

OpenAI GPT Latest — OpenAI added to the leaderboard

Qwen3.5 Plus 2026-04-20 — Qwen added to the leaderboard

Qwen3.6 Flash — Qwen added to the leaderboard

Qwen3.6 35B A3B — Qwen added to the leaderboard

Qwen3.6 Max Preview — Qwen added to the leaderboard

Qwen3.6 27B — Qwen added to the leaderboard

GPT-5.5 Pro — OpenAI added to the leaderboard

GPT-5.5 — OpenAI added to the leaderboard

DeepSeek V4 Pro — DeepSeek added to the leaderboard

DeepSeek V4 Flash — DeepSeek added to the leaderboard

DeepSeek V4 Pro — scored 90.1% on GPQA Diamond

DeepSeek V4 Pro — scored 37.7% on Humanity's Last Exam

DeepSeek V4 Pro — scored 87.5% on MMLU-Pro

DeepSeek V4 Pro — scored 80.6% on SWE-Bench Verified