GPT-6.2 Astra

  • New
  • Closed weights
  • Reasoning
  • Input: file · image · text
  • Released 2026-11
RankedEnough results for an index.
IndexAll categories, out of 100.79.5±3.1Median 50.7
CoverageIndex weight that has results.95%
SpeedOutput tokens per second.39/sMedian 76/s
First tokenSeconds to the first output token.8.6 sMedian 1.1 s
Input / 1MUS dollars per 1M input tokens.$8.00batch $4.00Median $0.60
Output / 1MUS dollars per 1M output tokens.$40.00batch $20.00 Batch price per 1M output tokens.Median $2.50
ContextMaximum tokens in one request.1.05MMedian 262K
EloLMArena rating and rank.1414±12 #141Median 1425

Details

From OpenRouter
API IDThe model ID on OpenRouter.
openai/gpt-6.2-astra
Max outputMaximum output tokens per request.
128K
Knowledge cutoffLast date of training data.
—
ToolsTool calls on OpenRouter.
Yes
JSON outputOutput follows a given JSON schema.
Yes
Cache read / 1MUS dollars per 1M cached tokens.
$0.40
Weights
—
RetiresDate OpenRouter removes the model.
—

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.15 counted74.710 / 221
CodingCode writing and repair.5 counted79.45 / 214
ReasoningLogic problems and puzzles.8 counted86.22 / 225
MultimodalTasks with images and text.1 counted70.7—
KnowledgeFacts and expert knowledge.5 counted79.92 / 266
MultilingualTasks in many languages.1 counted74.0—
InstructionTasks with strict prompt rules.None counted——
MathMath problems.3 counted76.43 / 171

Results

38 counted
BenchmarkResultThe score from the source.PlacePlace among results on this benchmark.
EuroEval PolishMultilingual68.5% ±0.31 / 227
GPQA diamond · Epoch AIKnowledge96.8% ±1.21 / 220
EuroEval GermanMultilingual68.1% ±0.51 / 208
OTIS Mock AIME 2024-2025Math100.0%1 / 203
EuroEval PortugueseMultilingual75.9% ±0.41 / 199
EuroEval FrenchMultilingual77.3% ±0.41 / 195
Graduate-Level Google-Proof Q&AKnowledge97.0%1 / 144
ARC-AGI-2 (semi-private)Reasoning95.8%1 / 137
ARC-AGI-1 (semi-private)Reasoning98.5%1 / 132
Artificial Analysis MMMU-ProMultimodal89.5%1 / 109
EuroEval SpanishMultilingual69.7% ±0.32 / 194
EuroEval DutchMultilingual74.6% ±0.32 / 192
EuroEval ItalianMultilingual78.1% ±0.32 / 191
GPQA Diamond · BenchLMKnowledge97.0%1 / 87
Mystery Game PuzzlesReasoningAbove fitFar above this model's index.95.0% ±2.21 / 83
EuroEval SwedishMultilingual77.4% ±0.43 / 213
SimpleQA VerifiedKnowledge78.5% ±1.32 / 94
EBR-benchReasoning92.4% ±4.91 / 33
Chess PuzzlesReasoning78.0% ±4.23 / 93
ARC-AGI-3 (semi-private)Reasoning70.3%1 / 30
Furniture AssemblyReasoning89.2% ±3.82 / 40
Terminal-Bench 4.0Agentic63.1% ±1.34 / 68
Terminal-Bench-Science 0.1Agentic77.1%1 / 15
Bug Hunt BenchCoding62.7%2 / 30
Vibe Code Bench v1.1Coding89.9% ±2.09 / 130
PostTrainBench v1.2Coding43.5%2 / 23
HealthBench ProfessionalKnowledge66.3%2 / 21
Terminal-Bench 4.1.0Agentic62.1% ±1.26 / 52
Agent Arena steerabilityAgentic14.07 / 59
FrontierCode 2.0Coding44.2%4 / 32
Agent Arena task outcomeAgentic12.410 / 59
MirrorCodeCoding67.8% ±11.78 / 40
Artificial Analysis GDP.pdfAgentic29.8%5 / 23
CWE-bench v2Agentic66.8%6 / 20
ARC-AGI-1 (public eval)Reasoning99.3%2 / 116
Terminal-Bench-ScienceScience71.4% ±5.41 / 48
MysteryMechanismScience53.2% ±3.41 / 34
ARC-AGI-2 (public eval)Reasoning97.5%6 / 127
FrontierMath-ErdosMath10.3% ±3.71 / 16
BioMysteryBenchScience80.7% ±0.43 / 34
SpireClimbOther41.3% ±4.21 / 11
HLAB-Vals-1.1Legal16.3% ±3.112 / 42
Legal Research BenchLegal42.8% ±3.431 / 98
Tax Agent BenchFinance21.8% ±2.854 / 89
Harvey LAB-AA v1.1Agentic15.0%—

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseEpoch AICC BYARC Prize FoundationTerms restrict reuseVals AINo licence statedEuroEvalMITTerminal-BenchNo licence statedLMArenaCC BY 4.0models.devMITCheaper InferenceNo licence stated

History

13 releases
Index79.0 → 79.5
18 Nov 202621 Nov 2026
Output / 1M$40.00 → $40.00
18 Nov 202621 Nov 2026

Questions

How much does GPT-6.2 Astra cost?
$8.00 per 1M input tokens, $40.00 per 1M output tokens.
How good is GPT-6.2 Astra?
Its index is 79.5 out of 100. That is place 4 of 282 ranked models.
What is GPT-6.2 Astra best at?
Knowledge: 79.9 out of 100, place 2 of 266.
How fast is GPT-6.2 Astra?
39 output tokens per second, 8.6 s to the first token.
What is the context window of GPT-6.2 Astra?
1.05M tokens.
Is GPT-6.2 Astra open source?
No. Its weights are not published.
When was GPT-6.2 Astra released?
November 2026.

Badge

GPT-6.2 Astra on Inference360: place 4 of 282, index 79.5

Compare with

Gemini 4 Carbon80.5Claude Opus 5.577.1Gemini 4 Neon81.9Gemini 4 Argon76.8GPT-6 Astra76.5

Lower price

Gemini 4 Carbon80.5 · $10.00/1MGemini 4 Argon76.8 · $10.00/1MGemini 4 Neon81.9 · $15.00/1M

More from OpenAI

All models
GPT-6.2 Sol76.0GPT-6.2 Sol ProUnrankedGPT-6.2 Astra ProUnrankedGPT-Live-1.1UnrankedGPT-6 CyberUnranked