DeepSeek-V4.1-Flash vs GPT-6 Astra
Wins 1 of 8 areas
Long documents
Wins 6 of 8 areas
Agents · Reasoning · Facts · Images and charts · Math · Following instructions
GPT-6 Astra is the stronger all-rounder.DeepSeek-V4.1-Flash is cheaper and better at long documents.
Scores updated · 38 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
- FactsGetting facts right instead of making them up03GPT-6 Astra3 of 3 tests
- MathCompetition and research-level math03GPT-6 Astra3 of 3 tests
- AgentsCarrying out multi-step tasks on its own12GPT-6 Astra2 of 3 tests
- Images and chartsUnderstanding pictures, charts and video01GPT-6 Astra1 of 2 tests · 1 tie
- Following instructionsDoing exactly what it is asked01GPT-6 Astra1 of 1 test
- Long documentsFinding answers in very long texts10DeepSeek-V4.1-Flash1 of 1 test
- CodingWriting and fixing software22Even2 each · 1 tie
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.
DeepSeek-V4.1-Flash costs 98% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where GPT-6 Astra pulls ahead
- Solves new puzzle games it has never seenARC-AGI-3+99points ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+51.3points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+45.2points ahead
Where DeepSeek-V4.1-Flash pulls ahead
- Fixes real GitHub issues, working as an agentLiveBench · Agentic Coding+20points ahead
- Reasons across sets of long documentsAA-LCR+3.3points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+2.2points ahead
Every test, side by side
All 38 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingEven
- LiveBench · Agentic CodingDeepSeek-V4.1-Flash by 2077.357.3+20
- LMArena · WebDevGPT-6 Astra by 155 rating points16191774+155 rating
- SciCodeGPT-6 Astra by 4.651.956.5+4.6
- Terminal-Bench 2.1DeepSeek-V4.1-Flash by 2.290.688.4+2.2
- LiveBench · Codingtie8080.4tie
AgentsGPT-6 Astra
- Terminal-Bench 4.0GPT-6 Astra by 32.326.859.1+32.3
- AA IT-Bench SREGPT-6 Astra by 1.746.948.6+1.7
- GDPValDeepSeek-V4.1-Flash by 1.45553.6+1.4
ReasoningGPT-6 Astra
- ARC-AGI-3GPT-6 Astra by 990.999.9+99
- CritPtGPT-6 Astra by 17.414.331.7+17.4
- ARC-AGI-2GPT-6 Astra by 16.978.195+16.9
- SimpleBenchGPT-6 Astra by 16.966.783.6+16.9
- Humanity's Last ExamGPT-6 Astra by 15.539.254.7+15.5
- LiveBench · ReasoningGPT-6 Astra by 686.792.7+6
- GPQA DiamondGPT-6 Astra by 5.290.996.1+5.2
FactsGPT-6 Astra
- AA-Omniscience · Non-hallucinationGPT-6 Astra by 45.23.548.7+45.2
- SimpleQA VerifiedGPT-6 Astra by 32.742.975.6+32.7
- AA-Omniscience · AccuracyGPT-6 Astra by 16.246.462.6+16.2
Images and chartsGPT-6 Astra
- MMMU-ProGPT-6 Astra by 9.97786.9+9.9
- LMArena · Visiontie12781280tie
MathGPT-6 Astra
- FrontierMath Tier 4GPT-6 Astra by 51.346.397.6+51.3
- FrontierMath Tiers 1-3 (v2)GPT-6 Astra by 22.571.293.7+22.5
- LiveBench · MathematicsGPT-6 Astra by 3.593.396.8+3.5
Long documentsDeepSeek-V4.1-Flash
- AA-LCRDeepSeek-V4.1-Flash by 3.38480.7+3.3
Following instructionsGPT-6 Astra
- LiveBench · Instruction FollowingGPT-6 Astra by 5.67075.6+5.6
Other results13 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.
- AA-OmniscienceGPT-6 Astra by 48.7-5.343.4+48.7
- Agents' Last ExamGPT-6 Astra by 27.531.859.3+27.5
- Terminal-Bench 4.0GPT-6 Astra by 26.531.257.7+26.5
- OSWorld 2.0GPT-6 Astra by 18.354.372.6+18.3
- AA IntelligenceGPT-6 Astra by 13.239.552.7+13.2
- livebench_languageGPT-6 Astra by 8.281.289.4+8.2
- HLE (with tools)DeepSeek-V4.1-Flash by 6.763.957.2+6.7
- vectara_factual_consistencyDeepSeek-V4.1-Flash by 4.595.891.3+4.5
- vectara_avg_summary_lengthGPT-6 Astra by 40.3 rating points108.2148.5+40.3 rating
- ARC-AGI-1GPT-6 Astra by 493.597.5+4
- livebench_data_analysisGPT-6 Astra by 3.779.383+3.7
- DeepSWE 1.1tie74.274.1tie
- vectara_answer_ratetie99.9100tie
Questions people ask
Which is better, DeepSeek-V4.1-Flash or GPT-6 Astra?
GPT-6 Astra wins six of the eight areas we test: agents, reasoning, facts, images and charts, math and following instructions. DeepSeek-V4.1-Flash wins long documents, and costs 98% less. They are level on coding.
Which is better for coding?
Neither. They win 2 coding tests each of the 5 both models report, and 1 is a tie.
Which is cheaper?
DeepSeek-V4.1-Flash costs $0.30 per million input tokens and $1.20 per million output tokens; GPT-6 Astra costs $10.00 and $50.00. That makes DeepSeek-V4.1-Flash about 98% cheaper for the same work.
How do you compare the two?
We use the 38 benchmark tests both models have published scores on. The verdict counts the 25 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 13 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.