# OpenAI GPT-6 Astra As of October 21, 2026, OpenAI GPT-6 Astra ranks #4 of 123 models overall with 83.0 ±4.1 on the LLMBenchmark Index. GPT-6 Astra ranks #1 of 75 in knowledge, its strongest category. OpenAI GPT-6 Astra costs $10.00 per million input tokens and $50.00 per million output tokens. GPT-6 Astra was released on September 4, 2026. Page: https://llmbenchmark.ai/models/gpt-6-astra Data from October 21, 2026. Every figure below names the board it was read from and the day. - Maker: OpenAI - Released: September 4, 2026 - Price per million input tokens: $10.00 - Price per million output tokens: $50.00 - Context window: 1.05M tokens - Overall rank: 4 of 123, by the LLMBenchmark Index - LLMBenchmark Index: 83.0 ±4.1 (90% margin, from 7 of 7 inputs) - Artificial Analysis Intelligence Index: 52.7 - Vibe Coding Index: 85.1 (from 5 of 6 coding benchmarks) Benchmark results (121 of 195): - LiveBench coding: 93.7% (source: LiveBench, read October 21, 2026) - LMArena WebDev: 1774 Elo (source: LMArena WebDev, read October 21, 2026) - Vals Index: 65.0% (source: Vals Index, read October 21, 2026) - AA Intelligence Index: 52.7 (source: Artificial Analysis (via OpenRouter), read October 21, 2026) - AA Coding Index: 76.9 (source: Artificial Analysis (via OpenRouter), read October 21, 2026) - AA Agentic Index: 51.0 (source: Artificial Analysis (via OpenRouter), read October 21, 2026) - Design Arena: websites: 1337 Elo (source: Design Arena, read October 21, 2026) - Design Arena: UI components: 1396 Elo (source: Design Arena, read October 21, 2026) - Design Arena: game dev: 1475 Elo (source: Design Arena, read October 21, 2026) - Design Arena: data viz: 1334 Elo (source: Design Arena, read October 21, 2026) - Design Arena: 3D: 1500 Elo (source: Design Arena, read October 21, 2026) - Design Arena: all code categories: 1379 Elo (source: Design Arena, read October 21, 2026) - Design Arena: SVG: 1431 Elo (source: Design Arena, read October 21, 2026) - Design Arena: ASCII art: 1353 Elo (source: Design Arena, read October 21, 2026) - Design Arena: agents building web apps: 1373 Elo (source: Design Arena, read October 21, 2026) - Design Arena: agents building full-stack apps: 1357 Elo (source: Design Arena, read October 21, 2026) - Design Arena: agents building mobile apps: 1315 Elo (source: Design Arena, read October 21, 2026) - Design Arena: agents building Godot games: 1348 Elo (source: Design Arena, read October 21, 2026) - Design Arena: PowerPoint slides: 1243 Elo (source: Design Arena, read October 21, 2026) - Design Arena: spreadsheets: 1361 Elo (source: Design Arena, read October 21, 2026) - LiveBench reasoning: 92.7% (source: LiveBench, read October 21, 2026) - LiveBench data analysis: 83.0% (source: LiveBench, read October 21, 2026) - LiveBench research: 85.2% (source: LiveBench, read October 21, 2026) - LiveBench browser: 84.0% (source: LiveBench, read October 21, 2026) - LiveBench long agentic: 82.4% (source: LiveBench, read October 21, 2026) - LiveBench design: 82.0% (source: LiveBench, read October 21, 2026) - Terminal-Bench 4 (Vals AI): 59.6% (source: Vals AI boards, read October 21, 2026) - Vibe Code Bench (Vals AI): 89.6% (source: Vals AI boards, read October 21, 2026) - Code migration (Vals AI): 67.7% (source: Vals AI boards, read October 21, 2026) - ProgramBench (Vals AI): 5.5% (source: Vals AI boards, read October 21, 2026) - SREBench (Vals AI): 56.9% (source: Vals AI boards, read October 21, 2026) - IOI (Vals AI): 100.0% (source: Vals AI boards, read October 21, 2026) - Terminal-Bench 2.1 (Vals AI): 87.3% (source: Vals AI boards, read October 21, 2026) - SWE-bench Multimodal (Vals AI): 43.3% (source: Vals AI boards, read October 21, 2026) - Vibe Code Bench 1-100 (Vals AI): 27.6% (source: Vals AI boards, read October 21, 2026) - CyberBench (Vals AI): 41.1% (source: Vals AI boards, read October 21, 2026) - Finance Agent (Vals AI): 53.5% (source: Vals AI boards, read October 21, 2026) - Tax Agent Bench (Vals AI): 20.7% (source: Vals AI boards, read October 21, 2026) - Legal Research Bench (Vals AI): 39.4% (source: Vals AI boards, read October 21, 2026) - Harvey's Legal Agent Benchmark (Vals AI): 7.1% (source: Vals AI boards, read October 21, 2026) - Excel Modeling Benchmark (Vals AI): 71.7% (source: Vals AI boards, read October 21, 2026) - BioMysteryBench (Vals AI): 79.3% (source: Vals AI boards, read October 21, 2026) - Terminal-Bench-Science (Vals AI): 62.9% (source: Vals AI boards, read October 21, 2026) - Vals Science Index: 65.1% (source: Vals AI boards, read October 21, 2026) - Time Horizon Index (Vals AI): 90.5% (source: Vals AI boards, read October 21, 2026) - CUA-bench (Vals AI): 19.2% (source: Vals AI boards, read October 21, 2026) - Robo-bench (Vals AI): 23.3% (source: Vals AI boards, read October 21, 2026) - MysteryMechanism (Vals AI): 53.1% (source: Vals AI boards, read October 21, 2026) - SAGE (Vals AI): 46.4% (source: Vals AI boards, read October 21, 2026) - MedCode (Vals AI): 48.5% (source: Vals AI boards, read October 21, 2026) - MedScribe (Vals AI): 87.9% (source: Vals AI boards, read October 21, 2026) - GPQA Diamond (Artificial Analysis run): 96.3% (source: Artificial Analysis, read October 21, 2026) - Humanity's Last Exam (Artificial Analysis run): 54.7% (source: Artificial Analysis, read October 21, 2026) - SciCode: 56.5% (source: Artificial Analysis, read October 21, 2026) - Long context reasoning (AA-LCR): 80.7% (source: Artificial Analysis, read October 21, 2026) - FrontierCode: 53.3% (source: Epoch AI benchmarking hub, read October 21, 2026) - FrontierSWE: 65.5% (source: Epoch AI benchmarking hub, read October 21, 2026) - GSO-Bench: 79.4% (source: Epoch AI benchmarking hub, read October 21, 2026) - WeirdML: 93.3% (source: Epoch AI benchmarking hub, read October 21, 2026) - MirrorCode: 46.7% (source: Epoch AI benchmarking hub, read October 21, 2026) - APEX-Agents: 64.7% (source: Epoch AI benchmarking hub, read October 21, 2026) - Remote Labor Index: 20.8% (source: Epoch AI benchmarking hub, read October 21, 2026) - BALROG: 68.3% (source: Epoch AI benchmarking hub, read October 21, 2026) - GPQA Diamond (Epoch AI run): 95.8% (source: Epoch AI benchmarking hub, read October 21, 2026) - ARC-AGI: 98.5% (source: Epoch AI benchmarking hub, read October 21, 2026) - ARC-AGI-2: 95.0% (source: Epoch AI benchmarking hub, read October 21, 2026) - Chess puzzles: 74.0% (source: Epoch AI benchmarking hub, read October 21, 2026) - Mystery game puzzles: 84.0% (source: Epoch AI benchmarking hub, read October 21, 2026) - EBR-bench: 76.2% (source: Epoch AI benchmarking hub, read October 21, 2026) - OTIS Mock AIME: 100.0% (source: Epoch AI benchmarking hub, read October 21, 2026) - FrontierMath, tiers 1 to 3: 93.7% (source: Epoch AI benchmarking hub, read October 21, 2026) - FrontierMath, tier 4: 97.6% (source: Epoch AI benchmarking hub, read October 21, 2026) - ProofBench: 99.0% (source: Epoch AI benchmarking hub, read October 21, 2026) - SimpleQA Verified: 75.6% (source: Epoch AI benchmarking hub, read October 21, 2026) - DTBench: 97.3% (source: Epoch AI benchmarking hub, read October 21, 2026) - LMCA: 64.4% (source: Epoch AI benchmarking hub, read October 21, 2026) - Furniture assembly: 80.0% (source: Epoch AI benchmarking hub, read October 21, 2026) - SpireClimb: 25.7% (source: Epoch AI benchmarking hub, read October 21, 2026) - ALE-Bench: 2951.3 (source: ALE-Bench (Sakana AI), read October 21, 2026) - Vending-Bench 2: $15,515 (source: Epoch AI benchmarking hub, read October 21, 2026) - Epoch Capabilities Index: 168.8 (source: Epoch AI benchmarking hub, read October 21, 2026) - LMArena Text: 1441 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: coding: 1489 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: hard prompts: 1463 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: math: 1465 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: instruction following: 1455 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: creative writing: 1420 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Vision: 1280 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: English: 1447 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: non-English: 1432 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: Chinese: 1489 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: German: 1459 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: Japanese: 1405 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: Korean: 1428 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: French: 1432 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: Spanish: 1422 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: Russian: 1438 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: Polish: 1422 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: multi-turn: 1447 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: longer queries: 1460 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: expert: 1489 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: hard prompts in English: 1464 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: software and IT: 1474 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: mathematical work: 1484 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: science: 1451 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: medicine and healthcare: 1425 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: legal and government: 1462 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: business and finance: 1439 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: writing and language: 1437 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Text: entertainment and media: 1427 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena WebDev: React: 1798 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena WebDev: HTML: 1775 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena WebDev: from an image: 1726 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Vision: English: 1280 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Vision: Chinese: 1361 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Vision: reading text: 1292 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Vision: diagrams: 1297 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Vision: homework: 1337 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - LMArena Vision: creative writing: 1286 Elo (source: LMArena (Hugging Face dataset), read October 21, 2026) - SciCode: 12.3% (source: Kaggle Benchmarks, read October 21, 2026) - FACTS Benchmark Suite: 75.7% (source: Kaggle Benchmarks, read October 21, 2026)