DeepSeek-V4-Flash vs Gemma 4 31B
Wins 7 of 7 areas
Coding · Agents · Reasoning · Facts · Math · Long documents · Following instructions
Wins 0 of 7 areas
—
DeepSeek-V4-Flash is the stronger all-rounder.Gemma 4 31B is cheaper.
Scores updated · 34 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.
- CodingWriting and fixing software70DeepSeek-V4-Flash7 of 8 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own41DeepSeek-V4-Flash4 of 5 tests
- ReasoningHard problems that need careful thinking30DeepSeek-V4-Flash3 of 3 tests
- MathCompetition and research-level math30DeepSeek-V4-Flash3 of 3 tests
- FactsGetting facts right instead of making them up21DeepSeek-V4-Flash2 of 3 tests
- Long documentsFinding answers in very long texts10DeepSeek-V4-Flash1 of 1 test
- Following instructionsDoing exactly what it is asked10DeepSeek-V4-Flash1 of 1 test
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.
Gemma 4 31B costs 68% 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 DeepSeek-V4-Flash pulls ahead
- Real work tasks from 44 professionsGDPVal+40.8points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+35.3points ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+27points ahead
Where Gemma 4 31B pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+6.7points ahead
- Finds the root cause of IT system incidentsAA IT-Bench SRE+5.8points ahead
Every test, side by side
All 34 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingDeepSeek-V4-Flash
- Terminal-Bench 2.1DeepSeek-V4-Flash by 35.378.743.4+35.3
- SWE-bench VerifiedDeepSeek-V4-Flash by 277952+27
- SWE-bench MultilingualDeepSeek-V4-Flash by 21.673.351.7+21.6
- SWE-bench ProDeepSeek-V4-Flash by 16.952.635.7+16.9
- LiveCodeBench v6DeepSeek-V4-Flash by 10.690.680+10.6
- LMArena · WebDevDeepSeek-V4-Flash by 65 rating points14301365+65 rating
- SciCodeDeepSeek-V4-Flash by 4.850.345.5+4.8
- Terminal-Bench Hardtie35.636.4tie
AgentsDeepSeek-V4-Flash
- GDPValDeepSeek-V4-Flash by 40.846.96.1+40.8
- τ-Bench V3 · BankingDeepSeek-V4-Flash by 24.639.414.8+24.6
- Terminal-Bench 4.0DeepSeek-V4-Flash by 12.112.10+12.1
- MCP AtlasDeepSeek-V4-Flash by 11.86957.2+11.8
- AA IT-Bench SREGemma 4 31B by 5.831.537.3+5.8
ReasoningDeepSeek-V4-Flash
- CritPtDeepSeek-V4-Flash by 15.216.61.4+15.2
- Humanity's Last ExamDeepSeek-V4-Flash by 1538.623.6+15
- GPQA DiamondDeepSeek-V4-Flash by 5.190.885.7+5.1
FactsDeepSeek-V4-Flash
- SimpleQA VerifiedDeepSeek-V4-Flash by 23.734.110.4+23.7
- AA-Omniscience · AccuracyDeepSeek-V4-Flash by 20.440.420+20.4
- AA-Omniscience · Non-hallucinationGemma 4 31B by 6.78.315+6.7
MathDeepSeek-V4-Flash
- HMMT Feb. 2026DeepSeek-V4-Flash by 17.694.877.2+17.6
- IMOAnswerBenchDeepSeek-V4-Flash by 13.988.474.5+13.9
- AIME 2026DeepSeek-V4-Flash by 6.695.889.2+6.6
Following instructionsDeepSeek-V4-Flash
- IFBenchDeepSeek-V4-Flash by 3.679.275.6+3.6
Other results10 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.
- NL2RepoDeepSeek-V4-Flash by 38.754.215.5+38.7
- τ²-Bench Telecom (AA run)DeepSeek-V4-Flash by 35.19559.9+35.1
- AA Agentic IndexDeepSeek-V4-Flash by 3541.76.7+35
- AA-OmniscienceDeepSeek-V4-Flash by 33.6-14.3-47.9+33.6
- Artificial Analysis Coding IndexDeepSeek-V4-Flash by 25.769.143.4+25.7
- AA IntelligenceDeepSeek-V4-Flash by 19.634.314.7+19.6
- HLE (with tools)DeepSeek-V4-Flash by 18.645.126.5+18.6
- Terminal-Bench 2.0DeepSeek-V4-Flash by 1456.942.9+14
- GPQA (unspecified)DeepSeek-V4-Flash by 4.288.584.3+4.2
- MMLU-ProDeepSeek-V4-Flash by 186.285.2+1
Questions people ask
Which is better, DeepSeek-V4-Flash or Gemma 4 31B?
DeepSeek-V4-Flash wins all seven areas where both have results: coding, agents, reasoning, facts, math, long documents and following instructions. Gemma 4 31B wins none, but costs 68% less.
Which is better for coding?
DeepSeek-V4-Flash. It wins 7 of the 8 coding tests both models report; Gemma 4 31B wins none, 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; Gemma 4 31B costs $0.17 and $0.40. That makes Gemma 4 31B about 68% cheaper for the same work.
How do you compare the two?
We use the 34 benchmark tests both models have published scores on. The verdict counts the 24 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 10 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.