Claude Haiku 4.5 vs Magistral Medium 1.2
Wins 4 of 7 areas
Coding · Agents · Long documents · Following instructions
Wins 2 of 7 areas
Reasoning · Images and charts
Claude Haiku 4.5 wins more areas, narrowly.Magistral Medium 1.2 is better at reasoning.
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.
- CodingWriting and fixing software30Claude Haiku 4.53 of 3 tests
- AgentsCarrying out multi-step tasks on its own20Claude Haiku 4.52 of 2 tests
- Long documentsFinding answers in very long texts10Claude Haiku 4.51 of 1 test
- Following instructionsDoing exactly what it is asked10Claude Haiku 4.51 of 1 test
- ReasoningHard problems that need careful thinking01Magistral Medium 1.21 of 3 tests · 2 ties
- Images and chartsUnderstanding pictures, charts and video01Magistral Medium 1.21 of 1 test
- FactsGetting facts right instead of making them up11Even1 each
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.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Claude Haiku 4.5 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+32.7points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+31.8points ahead
- Reasons across sets of long documentsAA-LCR+21.3points ahead
Where Magistral Medium 1.2 pulls ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+6.7points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+2.8points ahead
- Harder college exam questions with imagesMMMU-Pro+1.1points 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.
CodingClaude Haiku 4.5
- Terminal-Bench 2.1Claude Haiku 4.5 by 31.844.212.4+31.8
- Terminal-Bench HardClaude Haiku 4.5 by 14.427.312.9+14.4
- SciCodeClaude Haiku 4.5 by 342.239.2+3
AgentsClaude Haiku 4.5
- GDPValClaude Haiku 4.5 by 11.811.80+11.8
- τ-Bench V3 · BankingClaude Haiku 4.5 by 3.79.35.6+3.7
ReasoningMagistral Medium 1.2
- GPQA DiamondMagistral Medium 1.2 by 6.767.273.9+6.7
- CritPttie00.3tie
- Humanity's Last Examtie10.410.3tie
FactsEven
- AA-Omniscience · Non-hallucinationClaude Haiku 4.5 by 32.772.740+32.7
- AA-Omniscience · AccuracyMagistral Medium 1.2 by 2.81820.8+2.8
Images and chartsMagistral Medium 1.2
- MMMU-ProMagistral Medium 1.2 by 1.158.659.7+1.1
Following instructionsClaude Haiku 4.5
- IFBenchClaude Haiku 4.5 by 11.354.343+11.3
Other results5 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.
- Artificial Analysis Coding IndexClaude Haiku 4.5 by 22.643.921.3+22.6
- AA-OmniscienceClaude Haiku 4.5 by 22.3-4.4-26.7+22.3
- AA Agentic IndexClaude Haiku 4.5 by 8.410.31.9+8.4
- AA IntelligenceClaude Haiku 4.5 by 5.116.911.8+5.1
- τ²-Bench Telecom (AA run)Claude Haiku 4.5 by 2.754.752+2.7
Questions people ask
Which is better, Claude Haiku 4.5 or Magistral Medium 1.2?
Claude Haiku 4.5 wins four of the seven areas where both have results: coding, agents, long documents and following instructions. Magistral Medium 1.2 wins reasoning and images and charts. They are level on facts.
Which is better for coding?
Claude Haiku 4.5. It wins 3 of the 3 coding tests both models report; Magistral Medium 1.2 wins none.
How do you compare the two?
We use the 18 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 5 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.