DeepSeek-V3 vs DeepSeek-V3.1-Terminus
Wins 0 of 6 areas
—
Wins 6 of 6 areas
Coding · Agents · Reasoning · Facts · Long documents · Following instructions
DeepSeek-V3.1-Terminus is the stronger all-rounder.DeepSeek-V3 is cheaper.
Scores updated · 20 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.
- ReasoningHard problems that need careful thinking03DeepSeek-V3.1-Terminus3 of 3 tests
- AgentsCarrying out multi-step tasks on its own02DeepSeek-V3.1-Terminus2 of 3 tests · 1 tie
- FactsGetting facts right instead of making them up02DeepSeek-V3.1-Terminus2 of 2 tests
- Following instructionsDoing exactly what it is asked02DeepSeek-V3.1-Terminus2 of 2 tests
- CodingWriting and fixing software12DeepSeek-V3.1-Terminus2 of 3 tests
- Long documentsFinding answers in very long texts01DeepSeek-V3.1-Terminus1 of 1 test
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-V3 costs 10% 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-V3.1-Terminus pulls ahead
- Reasons across sets of long documentsAA-LCR+28.6points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+28points ahead
- Keeps track of context across a multi-turn chatMulti-Challenge+23points ahead
Where DeepSeek-V3 pulls ahead
- Code for real scientific research problemsSciCode+1points ahead
Every test, side by side
All 20 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingDeepSeek-V3.1-Terminus
- Terminal-Bench 2.1DeepSeek-V3.1-Terminus by 2816.944.9+28
- Terminal-Bench HardDeepSeek-V3.1-Terminus by 15.115.230.3+15.1
- SciCodeDeepSeek-V3 by 13938+1
AgentsDeepSeek-V3.1-Terminus
- τ-Bench V3 · BankingDeepSeek-V3.1-Terminus by 16.34.721+16.3
- GDPValDeepSeek-V3.1-Terminus by 10.6010.6+10.6
- Terminal-Bench 4.0tie00tie
ReasoningDeepSeek-V3.1-Terminus
- GPQA DiamondDeepSeek-V3.1-Terminus by 13.765.579.2+13.7
- Humanity's Last ExamDeepSeek-V3.1-Terminus by 11.74.716.4+11.7
- CritPtDeepSeek-V3.1-Terminus by 1.701.7+1.7
FactsDeepSeek-V3.1-Terminus
- AA-Omniscience · Non-hallucinationDeepSeek-V3.1-Terminus by 11.114.125.2+11.1
- AA-Omniscience · AccuracyDeepSeek-V3.1-Terminus by 2.225.427.7+2.2
Long documentsDeepSeek-V3.1-Terminus
- AA-LCRDeepSeek-V3.1-Terminus by 28.640.769.3+28.6
Following instructionsDeepSeek-V3.1-Terminus
- Multi-ChallengeDeepSeek-V3.1-Terminus by 2331.454.4+23
- IFBenchDeepSeek-V3.1-Terminus by 164157+16
Other results6 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-V3.1-Terminus by 58.627.586.1+58.6
- LiveCodeBenchDeepSeek-V3.1-Terminus by 25.749.274.9+25.7
- Artificial Analysis Coding IndexDeepSeek-V3.1-Terminus by 20.52343.5+20.5
- AA-OmniscienceDeepSeek-V3.1-Terminus by 14.3-40.7-26.4+14.3
- τ²-Bench Telecom (AA run)DeepSeek-V3 by 1047.137.1+10
- AA IntelligenceDeepSeek-V3.1-Terminus by 5.19.714.8+5.1
Questions people ask
Which is better, DeepSeek-V3 or DeepSeek-V3.1-Terminus?
DeepSeek-V3.1-Terminus wins all six areas where both have results: coding, agents, reasoning, facts, long documents and following instructions. DeepSeek-V3 wins none, but costs 10% less.
Which is better for coding?
DeepSeek-V3.1-Terminus. It wins 2 of the 3 coding tests both models report; DeepSeek-V3 wins 1.
Which is cheaper?
DeepSeek-V3 costs $0.24 per million input tokens and $0.90 per million output tokens; DeepSeek-V3.1-Terminus costs $0.27 and $1.00. That makes DeepSeek-V3 about 10% cheaper for the same work.
How do you compare the two?
We use the 20 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 6 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.