- Score
- 9.4 (+0.17 points)
- Total Cost (Current Price)
- $0.183 · 94.6% cheaper
- Response Time (avg)
- 6.24s · 71.1% faster
Meta
Release: 2026-10-13
Tested on: 2026-10-13 00:23
meta/muse-wonder::medium
~1T total (~50B active)MoEClosedEstimated
Summary
Muse Wonder scores 9.2 on AI BENCHY and ranks #30. It has 9.1 reliability, a 76.8% pass rate, $3.397 total cost, and 21.58s average response time.
What makes Muse Wonder unique: It stands out most in Trivia, where it ranks #2, while Anti-AI Tricks is its weakest area at #16.
Recommended alternatives
We recommend choosing one of these models instead:
- Score
- 9.1 (-0.07 points)
- Total Cost (Current Price)
- $0.140 · 95.9% cheaper
- Response Time (avg)
- 16.58s · 23.2% faster
- Score
- 9.6 (+0.36 points)
- Total Cost (Current Price)
- $0.460 · 86.5% cheaper
- Response Time (avg)
- 7.15s · 66.9% faster
Model facts
Researched on 2026-10-13
- Parameters
- ~1T total (~50B active)
- Architecture
- MoE
- Availability
- Closed
- License
- -
Best estimate from public evidence; the vendor did not disclose every value. Meta does not disclose parameter counts or dense-versus-MoE architecture for this exact checkpoint. The 1T total / 50B active MoE values are a provisional, low-confidence catalog estimate carried from the Muse Spark family; they are not vendor-reported or measured Muse Wonder specifications. Available through Meta Model API and OpenRouter with no official public weights or checkpoint license. Wonder Full is a separate subscription-only model, not the max effort of this API checkpoint. Supports minimal, low, medium, high, xhigh and max reasoning; none is explicitly unsupported. Meta keeps raw reasoning private; OpenRouter returns reasoning summaries and provider-reported reasoning-token usage.
9.2
Consistency
9.6
9.1
$3.397
Total Output Tokens
121,546
Total Input Tokens
241,489
Input Price
$4.000 / 1M
Output Price
$20.000 / 1M
Cache Read Price
$0.200 / 1M
Cache Write Price
N/A
Cache prices apply to input tokens. Reads reuse cached prompts; writes store them and can cost extra. Output tokens use the output price.
Flaky tests
1
Flaky tests had mixed outcomes across runs (at least one pass and one fail).
Hamster playing table tennis
Prompt: Create a detailed SVG illustration of a hamster playing table tennis.
#30 Muse Wonder
medium- Cost
- $0.208
- Time
- 64.8s
- Tokens
- 10,444 tok
Charts
Choose the first model, then click a second model to open a side-by-side page.
Top Models by Score
Score vs Total Cost
Response Time (avg)
Score vs Response Time (avg)
Total Output Tokens
Score vs Total Output Tokens
Quick Compare
Muse WondermediumvsClaude Opus 5.5mediumMuse WondermediumvsGPT-5.5mediumMuse WondermediumvsSeed 2.1 TurbolowMuse WondermediumvsClaude Sonnet 5.5maxMuse WondermediumvsClaude Opus 5mediumMuse WondermediumvsQwen3.8 MAXmediumMuse WondermediumvsClaude Opus 5highMuse WondermediumvsQwen3.8 Max (0902)highMuse WondermediumvsGemini 3.7 FlashlowMuse WondermediumvsGPT-5.3-Codexmedium
Category Breakdown
| Category | Score | Consistency | Tests Correct |
|---|---|---|---|
| Agentic | 10.0 | 10.0 | |
| Anti-AI Tricks | 8.3 | 10.0 | |
| Coding | 10.0 | 10.0 | |
| Combined | 10.0 | 10.0 | |
| Data parsing and extraction | 10.0 | 10.0 | |
| Domain specific | 5.3 | 10.0 | |
| General Intelligence | 10.0 | 10.0 | |
| Instructions following | 6.6 | 10.0 | |
| Puzzle Solving | 7.7 | 10.0 | |
| Tool Calling | 10.0 | 10.0 | |
| Trivia | 4.7 | 1.6 |