Claude Sonnet 5 vs Muse Spark Open
Wins 3 of 7 areas
Agents · Reasoning · Long documents
Wins 3 of 7 areas
Facts · Images and charts · Math
The two are evenly matched.Claude Sonnet 5 is better at agents and reasoning; Muse Spark Open at facts and math.
Scores updated · 16 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.
- FactsGetting facts right instead of making them up02Muse Spark Open2 of 3 tests · 1 tie
- MathCompetition and research-level math02Muse Spark Open2 of 2 tests
- Images and chartsUnderstanding pictures, charts and video01Muse Spark Open1 of 1 test
- AgentsCarrying out multi-step tasks on its own20Claude Sonnet 52 of 2 tests
- ReasoningHard problems that need careful thinking21Claude Sonnet 52 of 3 tests
- Long documentsFinding answers in very long texts10Claude Sonnet 51 of 1 test
- CodingWriting and fixing software11Even1 each
Images and charts and long documents 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 Spark Open costs 54% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Muse Spark Open pulls ahead
- Short factual questions, answered correctlySimpleQA Verified+22.9points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+15.3points ahead
- Unpublished advanced math problemsFrontierMath Tiers 1-3 (v2)+4.2points ahead
Where Claude Sonnet 5 pulls ahead
- Complex command-line tasks across many fieldsTerminal-Bench 4.0+6points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+4.1points ahead
- Reasons across sets of long documentsAA-LCR+3.7points ahead
Every test, side by side
All 16 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingEven
- SciCodeMuse Spark Open by 4.154.358.4+4.1
- Terminal-Bench 2.1Claude Sonnet 5 by 4.180.576.4+4.1
AgentsClaude Sonnet 5
- Terminal-Bench 4.0Claude Sonnet 5 by 614.18.1+6
- GDPValClaude Sonnet 5 by 3.148.245.1+3.1
ReasoningClaude Sonnet 5
- Humanity's Last ExamMuse Spark Open by 2.641.343.9+2.6
- GPQA DiamondClaude Sonnet 5 by 2.191.189+2.1
- CritPtClaude Sonnet 5 by 216.914.9+2
FactsMuse Spark Open
- SimpleQA VerifiedMuse Spark Open by 22.933.756.6+22.9
- AA-Omniscience · Non-hallucinationMuse Spark Open by 15.360.675.9+15.3
- AA-Omniscience · Accuracytie4039.3tie
Images and chartsMuse Spark Open
- MMMU-ProMuse Spark Open by 3.677.380.9+3.6
MathMuse Spark Open
- FrontierMath Tiers 1-3 (v2)Muse Spark Open by 4.265.669.8+4.2
- FrontierMath Tier 4Muse Spark Open by 2.429.331.7+2.4
Other results2 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-OmniscienceMuse Spark Open by 8.216.424.6+8.2
- AA Intelligencetie38.238.9tie
Questions people ask
Which is better, Claude Sonnet 5 or Muse Spark Open?
Claude Sonnet 5 and Muse Spark Open each win three of the seven areas where both have results. Claude Sonnet 5 wins agents, reasoning and long documents; Muse Spark Open wins facts, images and charts and math. Muse Spark Open costs 54% less. They are level on coding.
Which is better for coding?
Neither. They win 1 coding test each of the 2 both models report.
Which is cheaper?
Claude Sonnet 5 costs $2.00 per million input tokens and $10.00 per million output tokens; Muse Spark Open costs $1.25 and $4.25. That makes Muse Spark Open about 54% cheaper for the same work.
How do you compare the two?
We use the 16 benchmark tests both models have published scores on. The verdict counts the 14 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 2 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.