DeepSeek-V4-Pro vs Qwen3.8 27B
Wins 2 of 7 areas
Reasoning · Math
Wins 4 of 7 areas
Coding · Agents · Long documents · Following instructions
Qwen3.8 27B wins more areas, narrowly.DeepSeek-V4-Pro is better at reasoning.
Scores updated · 35 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.
- Following instructionsDoing exactly what it is asked02Qwen3.8 27B2 of 2 tests
- CodingWriting and fixing software45Qwen3.8 27B5 of 9 tests
- AgentsCarrying out multi-step tasks on its own23Qwen3.8 27B3 of 5 tests
- Long documentsFinding answers in very long texts01Qwen3.8 27B1 of 1 test
- ReasoningHard problems that need careful thinking32DeepSeek-V4-Pro3 of 5 tests
- MathCompetition and research-level math10DeepSeek-V4-Pro1 of 1 test
- FactsGetting facts right instead of making them up11Even1 each
Math and long documents 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.
Qwen3.8 27B costs 34% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Qwen3.8 27B pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+63.8points ahead
- Fixes real GitHub issues, working as an agentLiveBench · Agentic Coding+18.8points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+17.9points ahead
Where DeepSeek-V4-Pro pulls ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+27.4points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+21.8points ahead
- Complex command-line tasks across many fieldsTerminal-Bench 4.0+9points ahead
Every test, side by side
All 35 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingQwen3.8 27B
- LiveBench · Agentic CodingQwen3.8 27B by 18.842.661.4+18.8
- Terminal-Bench 2.1Qwen3.8 27B by 15.86479.8+15.8
- LMArena · WebDevQwen3.8 27B by 126 rating points14641590+126 rating
- SWE-bench ProQwen3.8 27B by 6.355.461.7+6.3
- LiveBench · CodingQwen3.8 27B by 5.77075.7+5.7
- SciCodeDeepSeek-V4-Pro by 4.250.846.6+4.2
- SWE-bench VerifiedDeepSeek-V4-Pro by 2.880.677.8+2.8
- SWE-bench MultilingualDeepSeek-V4-Pro by 2.476.273.8+2.4
- LiveCodeBench v6DeepSeek-V4-Pro by 2.292.590.3+2.2
AgentsQwen3.8 27B
- BrowseCompDeepSeek-V4-Pro by 21.883.461.6+21.8
- τ-Bench V3 · BankingQwen3.8 27B by 17.930.148+17.9
- GDPValQwen3.8 27B by 13.232.846+13.2
- Terminal-Bench 4.0DeepSeek-V4-Pro by 914.65.6+9
- MCP AtlasQwen3.8 27B by 3.773.677.3+3.7
ReasoningDeepSeek-V4-Pro
- SimpleBenchQwen3.8 27B by 9.350.960.2+9.3
- CritPtDeepSeek-V4-Pro by 7.512.95.4+7.5
- Humanity's Last ExamDeepSeek-V4-Pro by 3.637.533.9+3.6
- LiveBench · ReasoningDeepSeek-V4-Pro by 2.782.780+2.7
- GPQA DiamondQwen3.8 27B by 1.788.890.5+1.7
FactsEven
- AA-Omniscience · Non-hallucinationQwen3.8 27B by 63.85.969.7+63.8
- AA-Omniscience · AccuracyDeepSeek-V4-Pro by 27.44315.6+27.4
Following instructionsQwen3.8 27B
- LiveBench · Instruction FollowingQwen3.8 27B by 10.362.472.7+10.3
- IFBenchQwen3.8 27B by 376.579.5+3
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.
- AA Agentic IndexQwen3.8 27B by 18.827.746.5+18.8
- Agents' Last ExamQwen3.8 27B by 17.225.742.9+17.2
- Toolathlon VerifiedQwen3.8 27B by 11.255.967.1+11.2
- Artificial Analysis Coding IndexQwen3.8 27B by 8.759.468.1+8.7
- Terminal-Bench 4.0DeepSeek-V4-Pro by 6.812.45.6+6.8
- NL2RepoQwen3.8 27B by 3.838.542.3+3.8
- livebench_languageDeepSeek-V4-Pro by 3.778.174.3+3.7
- AA IntelligenceQwen3.8 27B by 3.330.433.7+3.3
- livebench_data_analysisQwen3.8 27B by 2.174.576.6+2.1
- AA-Omnisciencetie-10.7-10tie
Questions people ask
Which is better, DeepSeek-V4-Pro or Qwen3.8 27B?
Qwen3.8 27B wins four of the seven areas where both have results: coding, agents, long documents and following instructions. DeepSeek-V4-Pro wins reasoning and math. They are level on facts.
Which is better for coding?
Qwen3.8 27B. It wins 5 of the 9 coding tests both models report; DeepSeek-V4-Pro wins 4.
Which is cheaper?
DeepSeek-V4-Pro costs $1.32 per million input tokens and $3.96 per million output tokens; Qwen3.8 27B costs $0.50 and $3.00. That makes Qwen3.8 27B about 34% cheaper for the same work.
How do you compare the two?
We use the 35 benchmark tests both models have published scores on. The verdict counts the 25 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.