DeepSeek-V4-Flash vs GPT-5.4
Wins 0 of 7 areas
—
Wins 5 of 7 areas
Coding · Reasoning · Facts · Math · Long documents
GPT-5.4 is the stronger all-rounder.DeepSeek-V4-Flash is cheaper.
Scores updated · 50 tests both models report · How we compare
Where each one wins
Tests won in each of the seven areas where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- MathCompetition and research-level math06GPT-5.46 of 6 tests
- CodingWriting and fixing software16GPT-5.46 of 8 tests · 1 tie
- ReasoningHard problems that need careful thinking05GPT-5.45 of 5 tests
- FactsGetting facts right instead of making them up02GPT-5.42 of 3 tests · 1 tie
- Long documentsFinding answers in very long texts01GPT-5.41 of 1 test
- AgentsCarrying out multi-step tasks on its own22Even2 each · 1 tie
- Following instructionsDoing exactly what it is asked11Even1 each
Long documents rests on a single test.
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-Flash costs 90% 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-5.4 pulls ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+24.6points ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+22points ahead
- Unpublished advanced math problemsFrontierMath Tiers 1-3 (v2)+21.1points ahead
Every test, side by side
All 50 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGPT-5.4
- Terminal-Bench HardGPT-5.4 by 2235.657.6+22
- LiveBench · Agentic CodingGPT-5.4 by 746.853.8+7
- SciCodeGPT-5.4 by 6.350.356.6+6.3
- SWE-bench ProGPT-5.4 by 5.152.657.7+5.1
- LMArena · WebDevGPT-5.4 by 35 rating points14301465+35 rating
- LiveBench · CodingGPT-5.4 by 2.57577.5+2.5
- SWE-bench MultilingualDeepSeek-V4-Flash by 1.673.371.7+1.6
- Terminal-Bench 2.1tie78.778.3tie
AgentsEven
- BrowseCompGPT-5.4 by 9.573.282.7+9.5
- GDPValDeepSeek-V4-Flash by 9.546.937.4+9.5
- AA IT-Bench SREGPT-5.4 by 331.534.5+3
- MCP AtlasDeepSeek-V4-Flash by 1.86967.2+1.8
- τ-Bench V3 · Bankingtie39.439.6tie
ReasoningGPT-5.4
- ARC-AGI-2GPT-5.4 by 11.961.473.3+11.9
- CritPtGPT-5.4 by 6.816.623.4+6.8
- Humanity's Last ExamGPT-5.4 by 5.138.643.7+5.1
- LiveBench · ReasoningGPT-5.4 by 1.586.688.1+1.5
- GPQA DiamondGPT-5.4 by 1.290.892+1.2
FactsGPT-5.4
- SimpleQA VerifiedGPT-5.4 by 1134.145.1+11
- AA-Omniscience · AccuracyGPT-5.4 by 10.540.450.9+10.5
- AA-Omniscience · Non-hallucinationtie8.38.3tie
MathGPT-5.4
- FrontierMath Tier 4GPT-5.4 by 24.624.449+24.6
- FrontierMath Tiers 1-3 (v2)GPT-5.4 by 21.157.578.6+21.1
- LiveBench · MathematicsGPT-5.4 by 7.486.894.2+7.4
- AIME 2026GPT-5.4 by 3.495.899.2+3.4
- IMOAnswerBenchGPT-5.4 by 388.491.4+3
- HMMT Feb. 2026GPT-5.4 by 2.994.897.7+2.9
Following instructionsEven
- IFBenchDeepSeek-V4-Flash by 5.379.273.9+5.3
- LiveBench · Instruction FollowingGPT-5.4 by 4.765.570.2+4.7
Other results20 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.
- GDPval-AA (Elo)GPT-5.4 by 279 rating points13951674+279 rating
- Apex (Pass@1)GPT-5.4 by 21.13354.1+21.1
- AA-OmniscienceGPT-5.4 by 20.1-14.35.8+20.1
- Terminal-Bench 2.0GPT-5.4 by 18.256.975.1+18.2
- ToolathlonDeepSeek-V4-Flash by 15.770.354.6+15.7
- NL2RepoDeepSeek-V4-Flash by 12.954.241.3+12.9
- Codeforces (Rating)GPT-5.4 by 116 rating points30523168+116 rating
- CyberGymDeepSeek-V4-Flash by 10.476.766.3+10.4
- τ²-Bench Telecom (AA run)DeepSeek-V4-Flash by 7.99587.1+7.9
- Apex Shortlist (Pass@1)DeepSeek-V4-Flash by 7.685.778.1+7.6
- HLE (with tools)GPT-5.4 by 745.152.1+7
- AA IntelligenceGPT-5.4 by 4.734.339+4.7
- ARC-AGI-1GPT-5.4 by 4.78993.7+4.7
- livebench_languageGPT-5.4 by 3.479.282.6+3.4
- AA Agentic IndexGPT-5.4 by 2.541.744.2+2.5
- DeepSWE 1.1DeepSeek-V4-Flash by 2.454.452+2.4
- Chinese SimpleQA (C-SimpleQA)DeepSeek-V4-Flash by 2.178.976.8+2.1
- Artificial Analysis Coding IndexGPT-5.4 by 269.171.1+2
- MMLU-ProGPT-5.4 by 1.386.287.5+1.3
- livebench_data_analysistie79.379.3tie
Questions people ask
Which is better, DeepSeek-V4-Flash or GPT-5.4?
GPT-5.4 wins five of the seven areas where both have results: coding, reasoning, facts, math and long documents. DeepSeek-V4-Flash wins none, but costs 90% less. They are level on agents and following instructions.
Which is better for coding?
GPT-5.4. It wins 6 of the 8 coding tests both models report; DeepSeek-V4-Flash wins 1, and 1 is a tie.
Which is cheaper?
DeepSeek-V4-Flash costs $0.44 per million input tokens and $1.32 per million output tokens; GPT-5.4 costs $2.50 and $15.00. That makes DeepSeek-V4-Flash about 90% cheaper for the same work.
How do you compare the two?
We use the 50 benchmark tests both models have published scores on. The verdict counts the 30 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 20 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.