Claude 3.5 Haiku vs Kimi Linear 48B A3B Instruct
Wins 2 of 4 areas
Reasoning · Following instructions
Wins 0 of 4 areas
—
Claude 3.5 Haiku wins more areas, narrowly.
Scores updated · 9 tests both models report · How we compare
Where each one wins
Tests won in each of the four 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.
Long documents and following instructions rest on a single test each.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Claude 3.5 Haiku pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+14.7points ahead
- Code for real scientific research problemsSciCode+7.5points ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+1.1points ahead
Where Kimi Linear 48B A3B Instruct pulls ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+9.1points ahead
Every test, side by side
All 9 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingEven
- Terminal-Bench HardKimi Linear 48B A3B Instruct by 9.12.311.4+9.1
- SciCodeClaude 3.5 Haiku by 7.527.419.9+7.5
ReasoningClaude 3.5 Haiku
- Humanity's Last ExamClaude 3.5 Haiku by 1.13.62.5+1.1
- GPQA Diamondtie40.841.2tie
Following instructionsClaude 3.5 Haiku
- IFBenchClaude 3.5 Haiku by 14.742.828.1+14.7
Other results3 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)Claude 3.5 Haiku by 24.624.60+24.6
- Artificial Analysis Coding IndexClaude 3.5 Haiku by 1.715.914.2+1.7
- AA IntelligenceClaude 3.5 Haiku by 1.68.97.3+1.6
Questions people ask
Which is better, Claude 3.5 Haiku or Kimi Linear 48B A3B Instruct?
Claude 3.5 Haiku wins two of the four areas where both have results: reasoning and following instructions. Kimi Linear 48B A3B Instruct wins none. They are level on coding and long documents.
Which is better for coding?
Neither. They win 1 coding test each of the 2 both models report.
How do you compare the two?
We use the 9 benchmark tests both models have published scores on. The verdict counts the 6 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 3 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.