GPT-5.4 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-5.4 is cheaper and better at long documents.
Scores updated · 42 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 thinking06GPT-6 Astra6 of 6 tests
- CodingWriting and fixing software04GPT-6 Astra4 of 5 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own04GPT-6 Astra4 of 4 tests
- MathCompetition and research-level math03GPT-6 Astra3 of 3 tests
- FactsGetting facts right instead of making them up13GPT-6 Astra3 of 4 tests
- Following instructionsDoing exactly what it is asked01GPT-6 Astra1 of 1 test
- Long documentsFinding answers in very long texts10GPT-5.41 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-5.4 costs 71% 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+99.7points ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+48.6points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+40.4points ahead
Where GPT-5.4 pulls ahead
- Reasons across sets of long documentsAA-LCR+1.3points ahead
Every test, side by side
All 42 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 312 rating points14641776+312 rating
- Terminal-Bench 2.1GPT-6 Astra by 10.178.388.4+10.1
- LiveBench · Agentic CodingGPT-6 Astra by 3.553.857.3+3.5
- LiveBench · CodingGPT-6 Astra by 2.977.580.4+2.9
- SciCodetie56.656.5tie
AgentsGPT-6 Astra
- GDPValGPT-6 Astra by 16.337.353.6+16.3
- AA IT-Bench SREGPT-6 Astra by 14.134.548.6+14.1
- BrowseCompGPT-6 Astra by 8.882.791.5+8.8
- τ-Bench V3 · BankingGPT-6 Astra by 1.839.641.4+1.8
ReasoningGPT-6 Astra
- ARC-AGI-3GPT-6 Astra by 99.70.299.9+99.7
- ARC-AGI-2GPT-6 Astra by 21.773.395+21.7
- Humanity's Last ExamGPT-6 Astra by 1143.754.7+11
- CritPtGPT-6 Astra by 8.323.431.7+8.3
- LiveBench · ReasoningGPT-6 Astra by 4.688.192.7+4.6
- GPQA DiamondGPT-6 Astra by 4.19296.1+4.1
FactsGPT-6 Astra
- AA-Omniscience · Non-hallucinationGPT-6 Astra by 40.48.348.7+40.4
- SimpleQA VerifiedGPT-6 Astra by 30.545.175.6+30.5
- AA-Omniscience · AccuracyGPT-6 Astra by 11.750.962.6+11.7
- Vectara HHEM hallucination ratelower is betterGPT-5.4 by 1.778.7+1.7
Images and chartsEven
- MMMU-ProGPT-6 Astra by 8.578.486.9+8.5
- LMArena · VisionGPT-5.4 by 22 rating points13031281+22 rating
MathGPT-6 Astra
- FrontierMath Tier 4GPT-6 Astra by 48.64997.6+48.6
- FrontierMath Tiers 1-3 (v2)GPT-6 Astra by 15.178.693.7+15.1
- LiveBench · MathematicsGPT-6 Astra by 2.694.296.8+2.6
Following instructionsGPT-6 Astra
- LiveBench · Instruction FollowingGPT-6 Astra by 5.470.275.6+5.4
Other results16 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.
- HiL-Bench (Tools-allowed)GPT-6 Astra by 55.69.765.3+55.6
- AA-OmniscienceGPT-6 Astra by 37.65.843.4+37.6
- DeepSWE 1.1GPT-6 Astra by 22.15274.1+22.1
- HealthBench Professional length-adjustedGPT-6 Astra by 16.648.164.7+16.6
- AA IntelligenceGPT-6 Astra by 13.73952.7+13.7
- HealthBench Hard length-adjustedGPT-6 Astra by 7.529.136.6+7.5
- AA Agentic IndexGPT-6 Astra by 7.344.251.5+7.3
- livebench_languageGPT-6 Astra by 6.882.689.4+6.8
- vectara_avg_summary_lengthGPT-6 Astra by 66.8 rating points81.7148.5+66.8 rating
- Artificial Analysis Coding IndexGPT-6 Astra by 5.871.176.9+5.8
- HealthBench length-adjustedGPT-6 Astra by 4.35458.3+4.3
- ARC-AGI-1GPT-6 Astra by 3.893.797.5+3.8
- livebench_data_analysisGPT-6 Astra by 3.779.383+3.7
- vectara_factual_consistencyGPT-5.4 by 1.79391.3+1.7
- HealthBench Consensus length-adjustedtie96.395.5tie
- vectara_answer_ratetie99.9100tie
Questions people ask
Which is better, GPT-5.4 or GPT-6 Astra?
GPT-6 Astra wins six of the eight areas we test: coding, agents, reasoning, facts, math and following instructions. GPT-5.4 wins long documents, and costs 71% less. They are level on images and charts.
Which is better for coding?
GPT-6 Astra. It wins 4 of the 5 coding tests both models report; GPT-5.4 wins none, and 1 is a tie.
Which is cheaper?
GPT-5.4 costs $2.50 per million input tokens and $15.00 per million output tokens; GPT-6 Astra costs $10.00 and $50.00. That makes GPT-5.4 about 71% cheaper for the same work.
How do you compare the two?
We use the 42 benchmark tests both models have published scores on. The verdict counts the 26 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 16 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.