DeepSeek-V3.1-Terminus vs DeepSeek-V4-Pro
Wins 0 of 6 areas
—
Wins 5 of 6 areas
Coding · Agents · Reasoning · Long documents · Following instructions
DeepSeek-V4-Pro is the stronger all-rounder.DeepSeek-V3.1-Terminus is cheaper.
Scores updated · 19 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 software03DeepSeek-V4-Pro3 of 3 tests
- AgentsCarrying out multi-step tasks on its own03DeepSeek-V4-Pro3 of 3 tests
- ReasoningHard problems that need careful thinking03DeepSeek-V4-Pro3 of 3 tests
- Long documentsFinding answers in very long texts01DeepSeek-V4-Pro1 of 1 test
- Following instructionsDoing exactly what it is asked01DeepSeek-V4-Pro1 of 1 test
- FactsGetting facts right instead of making them up11Even1 each
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-V3.1-Terminus costs 3% 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-Pro pulls ahead
- Real work tasks from 44 professionsGDPVal+22.3points ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+21.1points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+19.5points ahead
Where DeepSeek-V3.1-Terminus pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+19.3points ahead
Every test, side by side
All 19 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingDeepSeek-V4-Pro
- Terminal-Bench 2.1DeepSeek-V4-Pro by 19.144.964+19.1
- Terminal-Bench HardDeepSeek-V4-Pro by 15.930.346.2+15.9
- SciCodeDeepSeek-V4-Pro by 12.83850.8+12.8
AgentsDeepSeek-V4-Pro
- GDPValDeepSeek-V4-Pro by 22.310.632.9+22.3
- Terminal-Bench 4.0DeepSeek-V4-Pro by 14.6014.6+14.6
- τ-Bench V3 · BankingDeepSeek-V4-Pro by 9.12130.1+9.1
ReasoningDeepSeek-V4-Pro
- Humanity's Last ExamDeepSeek-V4-Pro by 21.116.437.5+21.1
- CritPtDeepSeek-V4-Pro by 11.21.712.9+11.2
- GPQA DiamondDeepSeek-V4-Pro by 9.679.288.8+9.6
FactsEven
- AA-Omniscience · Non-hallucinationDeepSeek-V3.1-Terminus by 19.325.25.9+19.3
- AA-Omniscience · AccuracyDeepSeek-V4-Pro by 15.327.743+15.3
Following instructionsDeepSeek-V4-Pro
- IFBenchDeepSeek-V4-Pro by 19.55776.5+19.5
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.
- τ²-Bench Telecom (AA run)DeepSeek-V4-Pro by 59.137.196.2+59.1
- AA Agentic IndexDeepSeek-V4-Pro by 18.88.927.7+18.8
- LiveCodeBenchDeepSeek-V4-Pro by 18.674.993.5+18.6
- Artificial Analysis Coding IndexDeepSeek-V4-Pro by 15.943.559.4+15.9
- AA-OmniscienceDeepSeek-V4-Pro by 15.7-26.4-10.7+15.7
- AA IntelligenceDeepSeek-V4-Pro by 15.614.830.4+15.6
Questions people ask
Which is better, DeepSeek-V3.1-Terminus or DeepSeek-V4-Pro?
DeepSeek-V4-Pro wins five of the six areas where both have results: coding, agents, reasoning, long documents and following instructions. DeepSeek-V3.1-Terminus wins none, but costs 3% less. They are level on facts.
Which is better for coding?
DeepSeek-V4-Pro. It wins 3 of the 3 coding tests both models report; DeepSeek-V3.1-Terminus wins none.
Which is cheaper?
DeepSeek-V3.1-Terminus costs $0.27 per million input tokens and $1.00 per million output tokens; DeepSeek-V4-Pro costs $0.43 and $0.87. That makes DeepSeek-V3.1-Terminus about 3% cheaper for the same work.
How do you compare the two?
We use the 19 benchmark tests both models have published scores on. The verdict counts the 13 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.