Kimi K2.5 vs Muse Glimmer
Wins 4 of 8 areas
Coding · Reasoning · Facts · Math
Wins 3 of 8 areas
Agents · Long documents · Following instructions
Kimi K2.5 wins more areas, narrowly.Muse Glimmer is cheaper and better at agents.
Scores updated · 26 tests both models report · How we compare
Where each one wins
Tests won in each of the eight areas we test. 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 thinking20Kimi K2.52 of 3 tests · 1 tie
- FactsGetting facts right instead of making them up20Kimi K2.52 of 2 tests
- CodingWriting and fixing software21Kimi K2.52 of 5 tests · 2 ties
- MathCompetition and research-level math10Kimi K2.51 of 1 test
- AgentsCarrying out multi-step tasks on its own13Muse Glimmer3 of 4 tests
- Long documentsFinding answers in very long texts01Muse Glimmer1 of 1 test
- Following instructionsDoing exactly what it is asked01Muse Glimmer1 of 1 test
- Images and chartsUnderstanding pictures, charts and video11Even1 each
Math, 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.
Muse Glimmer costs 53% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Kimi K2.5 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+16.2points ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+8.7points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+8.2points ahead
Where Muse Glimmer pulls ahead
- Multi-step tasks using many real software toolsMCP Atlas+11.7points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+9.3points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+6.8points ahead
Every test, side by side
All 26 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingKimi K2.5
- LMArena · WebDevKimi K2.5 by 76 rating points14361360+76 rating
- Terminal-Bench 2.1Muse Glimmer by 645.751.7+6
- SciCodeKimi K2.5 by 4.14944.9+4.1
- SWE-bench Verifiedtie76.876tie
- SWE-bench Protie50.751.2tie
AgentsMuse Glimmer
- MCP AtlasMuse Glimmer by 11.763.875.5+11.7
- τ-Bench V3 · BankingMuse Glimmer by 9.314.223.5+9.3
- GDPValKimi K2.5 by 2.717.214.5+2.7
- OSWorld-VerifiedMuse Glimmer by 2.663.365.9+2.6
ReasoningKimi K2.5
- Humanity's Last ExamKimi K2.5 by 8.730.722+8.7
- GPQA DiamondKimi K2.5 by 4.487.983.5+4.4
- CritPttie3.12.6tie
FactsKimi K2.5
- AA-Omniscience · Non-hallucinationKimi K2.5 by 16.234.318.1+16.2
- AA-Omniscience · AccuracyKimi K2.5 by 8.235.227+8.2
Images and chartsEven
- CharXiv (RQ)Muse Glimmer by 1.377.578.8+1.3
- MMMU-ProKimi K2.5 by 1.175.474.3+1.1
Following instructionsMuse Glimmer
- IFBenchMuse Glimmer by 6.870.277+6.8
Other results7 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.
- AA-OmniscienceKimi K2.5 by 25.5-7.3-32.9+25.5
- OmniDocBench 1.5Kimi K2.5 by 1388.875.8+13
- AA Agentic IndexKimi K2.5 by 11.221.710.5+11.2
- τ³-Bench BankingMuse Glimmer by 9.314.223.5+9.3
- AA IntelligenceKimi K2.5 by 623.517.5+6
- DeepSearchQA (F1)Kimi K2.5 by 2.577.174.6+2.5
- Artificial Analysis Coding IndexMuse Glimmer by 2.246.849+2.2
Questions people ask
Which is better, Kimi K2.5 or Muse Glimmer?
Kimi K2.5 wins four of the eight areas we test: coding, reasoning, facts and math. Muse Glimmer wins agents, long documents and following instructions, and costs 53% less. They are level on images and charts.
Which is better for coding?
Kimi K2.5. It wins 2 of the 5 coding tests both models report; Muse Glimmer wins 1, and 2 are ties.
Which is cheaper?
Kimi K2.5 costs $0.60 per million input tokens and $3.00 per million output tokens; Muse Glimmer costs $0.33 and $1.35. That makes Muse Glimmer about 53% cheaper for the same work.
How do you compare the two?
We use the 26 benchmark tests both models have published scores on. The verdict counts the 19 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 7 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.