GPT-4o vs Qwen3 30B A3B 2507 Thinking
Wins 2 of 6 areas
Coding · Facts
Wins 3 of 6 areas
Reasoning · Long documents · Following instructions
Qwen3 30B A3B 2507 Thinking wins more areas, narrowly.GPT-4o is better at facts.
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 thinking02Qwen3 30B A3B 2507 Thinking2 of 3 tests · 1 tie
- Long documentsFinding answers in very long texts01Qwen3 30B A3B 2507 Thinking1 of 1 test
- Following instructionsDoing exactly what it is asked01Qwen3 30B A3B 2507 Thinking1 of 1 test
- FactsGetting facts right instead of making them up20GPT-4o2 of 2 tests
- CodingWriting and fixing software21GPT-4o2 of 4 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own00Even0 each · 1 tie
Agents, 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.
Qwen3 30B A3B 2507 Thinking costs 87% 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 30B A3B 2507 Thinking pulls ahead
- Recent programming contest problemsLiveCodeBench v6+35.1points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+18.1points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+14.7points ahead
Where GPT-4o pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+48.6points ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+11.2points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+7.4points 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.
CodingGPT-4o
- LiveCodeBench v6Qwen3 30B A3B 2507 Thinking by 35.130.966+35.1
- SWE-bench VerifiedGPT-4o by 11.233.222+11.2
- Terminal-Bench HardGPT-4o by 38.35.3+3
- SciCodetie33.333tie
ReasoningQwen3 30B A3B 2507 Thinking
- GPQA DiamondQwen3 30B A3B 2507 Thinking by 18.152.670.7+18.1
- Humanity's Last ExamQwen3 30B A3B 2507 Thinking by 8.51.810.3+8.5
- CritPttie00.3tie
FactsGPT-4o
- AA-Omniscience · Non-hallucinationGPT-4o by 48.662.113.5+48.6
- AA-Omniscience · AccuracyGPT-4o by 7.423.716.3+7.4
Long documentsQwen3 30B A3B 2507 Thinking
- AA-LCRQwen3 30B A3B 2507 Thinking by 5.35661.3+5.3
Following instructionsQwen3 30B A3B 2507 Thinking
- IFBenchQwen3 30B A3B 2507 Thinking by 14.73650.7+14.7
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.
- AIME 2025Qwen3 30B A3B 2507 Thinking by 73.311.785+73.3
- AA-OmniscienceGPT-4o by 45.6-10.5-56.1+45.6
- MATH500 (Pass@1)Qwen3 30B A3B 2507 Thinking by 11.974.686.5+11.9
- IFEvalQwen3 30B A3B 2507 Thinking by 9.88190.8+9.8
- τ²-Bench (Retail)GPT-4o by 7.363.456.1+7.3
- MMLU-ProQwen3 30B A3B 2507 Thinking by 7.274.781.9+7.2
- AA IntelligenceQwen3 30B A3B 2507 Thinking by 2.57.39.8+2.5
- τ²-Bench Telecom (AA run)tie28.928.1tie
Questions people ask
Which is better, GPT-4o or Qwen3 30B A3B 2507 Thinking?
Qwen3 30B A3B 2507 Thinking wins three of the six areas where both have results: reasoning, long documents and following instructions. GPT-4o wins coding and facts. They are level on agents.
Which is better for coding?
GPT-4o. It wins 2 of the 4 coding tests both models report; Qwen3 30B A3B 2507 Thinking wins 1, and 1 is a tie.
Which is cheaper?
GPT-4o costs $5.00 per million input tokens and $15.00 per million output tokens; Qwen3 30B A3B 2507 Thinking costs $0.20 and $2.40. That makes Qwen3 30B A3B 2507 Thinking about 87% 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 12 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.