GPT-5.5 vs GPT-6 Astra
Wins 1 of 8 areas
Long documents
Wins 6 of 8 areas
Coding · Agents · Reasoning · Facts · Math · Following instructions
GPT-6 Astra is the stronger all-rounder.GPT-5.5 is cheaper and better at long documents.
Both rank among the ten best models we track in reasoning and math.
Scores updated · 47 tests both models report · How we compare
Where each one wins
Tests won in each of the eight areas we test. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- ReasoningHard problems that need careful thinking07GPT-6 Astra7 of 7 tests
- AgentsCarrying out multi-step tasks on its own05GPT-6 Astra5 of 5 tests
- FactsGetting facts right instead of making them up03GPT-6 Astra3 of 4 tests · 1 tie
- CodingWriting and fixing software13GPT-6 Astra3 of 5 tests · 1 tie
- MathCompetition and research-level math02GPT-6 Astra2 of 3 tests · 1 tie
- Following instructionsDoing exactly what it is asked01GPT-6 Astra1 of 1 test
- Long documentsFinding answers in very long texts10GPT-5.51 of 1 test
- Images and chartsUnderstanding pictures, charts and video11Even1 each
Long documents and following instructions rest on a single test each.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
GPT-5.5 costs 42% less for the same work.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where GPT-6 Astra pulls ahead
- Solves new puzzle games it has never seenARC-AGI-3+99.5points ahead
- Complex command-line tasks across many fieldsTerminal-Bench 4.0+44.5points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+37.7points ahead
Where GPT-5.5 pulls ahead
- Reasons across sets of long documentsAA-LCR+3.6points ahead
- Programming problems, refreshed regularlyLiveBench · Coding+1.8points ahead
Every test, side by side
All 47 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGPT-6 Astra
- LMArena · WebDevGPT-6 Astra by 289 rating points14871776+289 rating
- Terminal-Bench 2.1GPT-6 Astra by 4.184.388.4+4.1
- LiveBench · Agentic CodingGPT-6 Astra by 3.35457.3+3.3
- LiveBench · CodingGPT-5.5 by 1.882.280.4+1.8
- SciCodetie55.856.5tie
AgentsGPT-6 Astra
- Terminal-Bench 4.0GPT-6 Astra by 44.514.659.1+44.5
- GDPValGPT-6 Astra by 11.142.553.6+11.1
- BrowseCompGPT-6 Astra by 7.184.491.5+7.1
- AA IT-Bench SREGPT-6 Astra by 2.845.848.6+2.8
- τ-Bench V3 · BankingGPT-6 Astra by 2.43941.4+2.4
ReasoningGPT-6 Astra
- ARC-AGI-3GPT-6 Astra by 99.50.499.9+99.5
- SimpleBenchGPT-6 Astra by 14.66983.6+14.6
- ARC-AGI-2GPT-6 Astra by 108595+10
- Humanity's Last ExamGPT-6 Astra by 8.945.854.7+8.9
- CritPtGPT-6 Astra by 4.627.131.7+4.6
- LiveBench · ReasoningGPT-6 Astra by 389.792.7+3
- GPQA DiamondGPT-6 Astra by 2.693.596.1+2.6
FactsGPT-6 Astra
- AA-Omniscience · Non-hallucinationGPT-6 Astra by 37.71148.7+37.7
- SimpleQA VerifiedGPT-6 Astra by 12.66375.6+12.6
- AA-Omniscience · AccuracyGPT-6 Astra by 4.65862.6+4.6
- Vectara HHEM hallucination ratelower is bettertie9.38.7tie
Images and chartsEven
- MMMU-ProGPT-6 Astra by 779.986.9+7
- LMArena · VisionGPT-5.5 by 11 rating points12921281+11 rating
MathGPT-6 Astra
- FrontierMath Tier 4GPT-6 Astra by 25.172.597.6+25.1
- FrontierMath Tiers 1-3 (v2)GPT-6 Astra by 8.485.393.7+8.4
- LiveBench · Mathematicstie95.996.8tie
Following instructionsGPT-6 Astra
- LiveBench · Instruction FollowingGPT-6 Astra by 4.970.775.6+4.9
Other results19 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- Agents' Last ExamGPT-6 Astra by 32.726.659.3+32.7
- HiL-Bench (Tools-allowed)GPT-6 Astra by 25.639.765.3+25.6
- AA-OmniscienceGPT-6 Astra by 22.920.543.4+22.9
- PostTrainBenchGPT-6 Astra by 15.928.444.3+15.9
- AA IntelligenceGPT-6 Astra by 14.338.452.7+14.3
- AA Agentic IndexGPT-6 Astra by 14.237.351.5+14.2
- HealthBench Professional length-adjustedGPT-6 Astra by 12.951.864.7+12.9
- HealthBench ProfessionalGPT-6 Astra by 11.651.863.4+11.6
- DeepSWE 1.1GPT-6 Astra by 7.16774.1+7.1
- HealthBench Hard length-adjustedGPT-6 Astra by 5.131.536.6+5.1
- ARC-AGI-1GPT-6 Astra by 2.59597.5+2.5
- Artificial Analysis Coding IndexGPT-6 Astra by 274.976.9+2
- livebench_languageGPT-6 Astra by 287.489.4+2
- vectara_avg_summary_lengthGPT-6 Astra by 18.9 rating points129.6148.5+18.9 rating
- HealthBench length-adjustedGPT-6 Astra by 1.856.558.3+1.8
- livebench_data_analysisGPT-6 Astra by 1.481.683+1.4
- vectara_factual_consistencytie90.791.3tie
- HealthBench Consensus length-adjustedtie95.695.5tie
- vectara_answer_ratetie100100tie
Questions people ask
Which is better, GPT-5.5 or GPT-6 Astra?
GPT-6 Astra wins six of the eight areas we test: coding, agents, reasoning, facts, math and following instructions. GPT-5.5 wins long documents, and costs 42% less. They are level on images and charts.
Which is better for coding?
GPT-6 Astra. It wins 3 of the 5 coding tests both models report; GPT-5.5 wins 1, and 1 is a tie.
Which is cheaper?
GPT-5.5 costs $5.00 per million input tokens and $30.00 per million output tokens; GPT-6 Astra costs $10.00 and $50.00. That makes GPT-5.5 about 42% cheaper for the same work.
How do you compare the two?
We use the 47 benchmark tests both models have published scores on. The verdict counts the 28 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 19 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.