Claude Haiku 4.5 vs GPT-5.1 Codex mini
Wins 1 of 7 areas
Long documents
Wins 4 of 7 areas
Agents · Reasoning · Images and charts · Following instructions
GPT-5.1 Codex mini wins more areas, narrowly.Claude Haiku 4.5 is better at long documents.
Scores updated · 18 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.
- ReasoningHard problems that need careful thinking02GPT-5.1 Codex mini2 of 3 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own01GPT-5.1 Codex mini1 of 1 test
- Images and chartsUnderstanding pictures, charts and video01GPT-5.1 Codex mini1 of 1 test
- Following instructionsDoing exactly what it is asked01GPT-5.1 Codex mini1 of 1 test
- Long documentsFinding answers in very long texts10Claude Haiku 4.51 of 1 test
- CodingWriting and fixing software11Even1 each · 1 tie
- FactsGetting facts right instead of making them up11Even1 each
Agents, images and charts, 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.
GPT-5.1 Codex mini costs 63% 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 GPT-5.1 Codex mini pulls ahead
- Real work tasks from 44 professionsGDPVal+15.9points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+14.1points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+13.6points ahead
Where Claude Haiku 4.5 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+24.3points ahead
- Reasons across sets of long documentsAA-LCR+7.6points ahead
Every test, side by side
All 18 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingEven
- LMArena · WebDevClaude Haiku 4.5 by 86 rating points13291243+86 rating
- Terminal-Bench HardGPT-5.1 Codex mini by 627.333.3+6
- SciCodetie42.242.6tie
ReasoningGPT-5.1 Codex mini
- GPQA DiamondGPT-5.1 Codex mini by 14.167.281.3+14.1
- Humanity's Last ExamGPT-5.1 Codex mini by 8.110.418.5+8.1
- CritPttie00tie
FactsEven
- AA-Omniscience · Non-hallucinationClaude Haiku 4.5 by 24.372.748.4+24.3
- AA-Omniscience · AccuracyGPT-5.1 Codex mini by 5.21823.2+5.2
Images and chartsGPT-5.1 Codex mini
- MMMU-ProGPT-5.1 Codex mini by 10.458.669+10.4
Following instructionsGPT-5.1 Codex mini
- IFBenchGPT-5.1 Codex mini by 13.654.367.9+13.6
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.
- AIME 2025Claude Haiku 4.5 by 38.680.742.1+38.6
- AA Agentic IndexGPT-5.1 Codex mini by 28.410.338.7+28.4
- AA-OmniscienceClaude Haiku 4.5 by 12-4.4-16.4+12
- τ²-Bench Telecom (AA run)GPT-5.1 Codex mini by 8.254.762.9+8.2
- Artificial Analysis Coding IndexClaude Haiku 4.5 by 7.543.936.4+7.5
- AA IntelligenceGPT-5.1 Codex mini by 3.516.920.4+3.5
Questions people ask
Which is better, Claude Haiku 4.5 or GPT-5.1 Codex mini?
GPT-5.1 Codex mini wins four of the seven areas where both have results: agents, reasoning, images and charts and following instructions. Claude Haiku 4.5 wins long documents. They are level on coding and facts.
Which is better for coding?
Neither. They win 1 coding test each of the 3 both models report, and 1 is a tie.
Which is cheaper?
Claude Haiku 4.5 costs $1.00 per million input tokens and $5.00 per million output tokens; GPT-5.1 Codex mini costs $0.25 and $2.00. That makes GPT-5.1 Codex mini about 63% cheaper for the same work.
How do you compare the two?
We use the 18 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 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.