DeepSeek-V3.2 vs Qwen3 30B A3B 2507 Thinking
Wins 7 of 7 areas
Coding · Agents · Reasoning · Facts · Math · Long documents · Following instructions
Wins 0 of 7 areas
—
DeepSeek-V3.2 is the stronger all-rounder.
Scores updated · 24 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 software40DeepSeek-V3.24 of 4 tests
- AgentsCarrying out multi-step tasks on its own30DeepSeek-V3.23 of 3 tests
- ReasoningHard problems that need careful thinking30DeepSeek-V3.23 of 3 tests
- FactsGetting facts right instead of making them up20DeepSeek-V3.22 of 2 tests
- MathCompetition and research-level math20DeepSeek-V3.22 of 2 tests
- Long documentsFinding answers in very long texts10DeepSeek-V3.21 of 1 test
- Following instructionsDoing exactly what it is asked10DeepSeek-V3.21 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.
DeepSeek-V3.2 costs 73% 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 pulls ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+51.1points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+49.1points ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+30.3points ahead
Where Qwen3 30B A3B 2507 Thinking pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 24 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingDeepSeek-V3.2
- SWE-bench VerifiedDeepSeek-V3.2 by 51.173.122+51.1
- Terminal-Bench HardDeepSeek-V3.2 by 30.335.65.3+30.3
- LiveCodeBench v6DeepSeek-V3.2 by 17.383.366+17.3
- SciCodeDeepSeek-V3.2 by 5.938.933+5.9
AgentsDeepSeek-V3.2
- BrowseCompDeepSeek-V3.2 by 49.151.42.3+49.1
- τ-Bench V3 · BankingDeepSeek-V3.2 by 13.418.85.4+13.4
- GDPValDeepSeek-V3.2 by 9.89.80+9.8
ReasoningDeepSeek-V3.2
- Humanity's Last ExamDeepSeek-V3.2 by 14.324.610.3+14.3
- GPQA DiamondDeepSeek-V3.2 by 13.38470.7+13.3
- CritPtDeepSeek-V3.2 by 2.62.90.3+2.6
FactsDeepSeek-V3.2
- AA-Omniscience · AccuracyDeepSeek-V3.2 by 16.73316.3+16.7
- AA-Omniscience · Non-hallucinationDeepSeek-V3.2 by 3.817.313.5+3.8
MathDeepSeek-V3.2
- AIME 2026DeepSeek-V3.2 by 5.994.288.3+5.9
- HMMT Feb. 2026DeepSeek-V3.2 by 5.384.178.8+5.3
Following instructionsDeepSeek-V3.2
- IFBenchDeepSeek-V3.2 by 1060.750.7+10
Other results8 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-V3.2 by 62.590.628.1+62.5
- AA-OmniscienceDeepSeek-V3.2 by 33.6-22.5-56.1+33.6
- τ²-BenchDeepSeek-V3.2 by 31.380.349+31.3
- AIME25 no toolsDeepSeek-V3.2 by 17.689.371.7+17.6
- AA IntelligenceDeepSeek-V3.2 by 11.721.59.8+11.7
- AIME 2025DeepSeek-V3.2 by 8.193.185+8.1
- GPQA (unspecified)DeepSeek-V3.2 by 6.579.973.4+6.5
- MMLU-ProDeepSeek-V3.2 by 3.18581.9+3.1
Questions people ask
Which is better, DeepSeek-V3.2 or Qwen3 30B A3B 2507 Thinking?
DeepSeek-V3.2 wins all seven areas where both have results: coding, agents, reasoning, facts, math, long documents and following instructions. Qwen3 30B A3B 2507 Thinking wins none.
Which is better for coding?
DeepSeek-V3.2. It wins 4 of the 4 coding tests both models report; Qwen3 30B A3B 2507 Thinking wins none.
Which is cheaper?
DeepSeek-V3.2 costs $0.28 per million input tokens and $0.42 per million output tokens; Qwen3 30B A3B 2507 Thinking costs $0.20 and $2.40. That makes DeepSeek-V3.2 about 73% cheaper for the same work.
How do you compare the two?
We use the 24 benchmark tests both models have published scores on. The verdict counts the 16 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 8 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.