DeepSeek-V3.1 vs DeepSeek-V3.2-Exp
Wins 0 of 6 areas
—
Wins 5 of 6 areas
Coding · Agents · Reasoning · Long documents · Following instructions
DeepSeek-V3.2-Exp is the stronger all-rounder.
Scores updated · 31 tests both models report · How we compare
Where each one wins
Tests won in each of the six 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 software13DeepSeek-V3.2-Exp3 of 4 tests
- AgentsCarrying out multi-step tasks on its own02DeepSeek-V3.2-Exp2 of 2 tests
- ReasoningHard problems that need careful thinking01DeepSeek-V3.2-Exp1 of 3 tests · 2 ties
- Long documentsFinding answers in very long texts01DeepSeek-V3.2-Exp1 of 1 test
- Following instructionsDoing exactly what it is asked01DeepSeek-V3.2-Exp1 of 1 test
- FactsGetting facts right instead of making them up11Even1 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-V3.2-Exp costs 69% 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-V3.2-Exp pulls ahead
Where DeepSeek-V3.1 pulls ahead
- Code for real scientific research problemsSciCode+1.4points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+1.4points ahead
Every test, side by side
All 31 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingDeepSeek-V3.2-Exp
- Terminal-Bench HardDeepSeek-V3.2-Exp by 6.12531.1+6.1
- SWE-bench MultilingualDeepSeek-V3.2-Exp by 3.454.557.9+3.4
- SWE-bench VerifiedDeepSeek-V3.2-Exp by 1.86667.8+1.8
- SciCodeDeepSeek-V3.1 by 1.439.137.7+1.4
AgentsDeepSeek-V3.2-Exp
- GDPValDeepSeek-V3.2-Exp by 19.45.625+19.4
- BrowseCompDeepSeek-V3.2-Exp by 10.13040.1+10.1
ReasoningDeepSeek-V3.2-Exp
- GPQA DiamondDeepSeek-V3.2-Exp by 1.877.979.7+1.8
- CritPttie21.4tie
- Humanity's Last Examtie14.314.9tie
FactsEven
- AA-Omniscience · Non-hallucinationDeepSeek-V3.2-Exp by 1.717.519.2+1.7
- AA-Omniscience · AccuracyDeepSeek-V3.1 by 1.42927.6+1.4
- Vectara HHEM hallucination ratelower is bettertie5.55.3tie
Long documentsDeepSeek-V3.2-Exp
- AA-LCRDeepSeek-V3.2-Exp by 15.656.772.3+15.6
Following instructionsDeepSeek-V3.2-Exp
- IFBenchDeepSeek-V3.2-Exp by 12.641.554.1+12.6
Other results17 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.
- HMMT 2025DeepSeek-V3.2-Exp by 50.133.583.6+50.1
- AIME 2025DeepSeek-V3.2-Exp by 39.549.889.3+39.5
- LiveCodeBenchDeepSeek-V3.2-Exp by 17.756.474.1+17.7
- AA Agentic IndexDeepSeek-V3.2-Exp by 9.818.928.7+9.8
- Terminal-BenchDeepSeek-V3.2-Exp by 6.431.337.7+6.4
- Aider-PolyglotDeepSeek-V3.2-Exp by 6.168.474.5+6.1
- HMMT Feb. 2025DeepSeek-V3.2-Exp by 4.285.890+4.2
- SimpleQADeepSeek-V3.2-Exp by 3.793.497.1+3.7
- Artificial Analysis Coding IndexDeepSeek-V3.2-Exp by 3.629.733.3+3.6
- τ²-Bench Telecom (AA run)DeepSeek-V3.1 by 3.537.433.9+3.5
- AA IntelligenceDeepSeek-V3.2-Exp by 3.113.516.6+3.1
- vectara_answer_rateDeepSeek-V3.2-Exp by 2.194.596.6+2.1
- AA-OmniscienceDeepSeek-V3.1 by 1.4-29.6-31+1.4
- BrowseComp-ZHDeepSeek-V3.1 by 1.349.247.9+1.3
- MMLU-ProDeepSeek-V3.2-Exp by 1.383.785+1.3
- vectara_avg_summary_lengthtie63.764.6tie
- vectara_factual_consistencytie94.594.7tie
Questions people ask
Which is better, DeepSeek-V3.1 or DeepSeek-V3.2-Exp?
DeepSeek-V3.2-Exp wins five of the six areas where both have results: coding, agents, reasoning, long documents and following instructions. DeepSeek-V3.1 wins none. They are level on facts.
Which is better for coding?
DeepSeek-V3.2-Exp. It wins 3 of the 4 coding tests both models report; DeepSeek-V3.1 wins 1.
Which is cheaper?
DeepSeek-V3.1 costs $0.56 per million input tokens and $1.68 per million output tokens; DeepSeek-V3.2-Exp costs $0.28 and $0.42. That makes DeepSeek-V3.2-Exp about 69% cheaper for the same work.
How do you compare the two?
We use the 31 benchmark tests both models have published scores on. The verdict counts the 14 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 17 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.