vs
1:10
overall capability wins (0 ties across 11 scored). Muse Spark 1.2 leads overall.
- Reasoning1:4
- Coding0:3
- Arena0:1
- Tool use0:1
- Composite indices0:2
- Specs0:1
Verdict
Muse Spark 1.2 leads coding (widest gap: +175.3 on WebDev Elo); Muse Spark 1.2 ranks higher on LMArena (+72 Elo); Muse Spark 1.2 offers ~8.0x larger context (1.0M vs 131K).
- Muse Spark 1.2 leads coding (widest gap: +175.3 on WebDev Elo)
- Muse Spark 1.2 ranks higher on LMArena (+72 Elo)
- Muse Spark 1.2 offers ~8.0x larger context (1.0M vs 131K)
- Muse Glimmer ships open weights (Apache 2.0)
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- Muse Glimmer
- -
- Muse Spark 1.2
- $2.31
Price per 1M tokens
Full comparison
Organization
Meta
Meta
Family
Llama / Muse
Llama / Muse
License
Apache 2.0
Proprietary
Open weights
Yes
No
Release
Aug 10, 2026
Aug 5, 2026
Knowledge cutoff
2026-01-04
-
API / provider
No primary API price listed
Meta
Modalities
text, image → text
text, image, video, audio, pdf → text
Specs
Context window
131K
1.0M
Max output
-
131K
Parameters
30B
—
Pricing
Input $/1M
—
$1.25
Output $/1M
—
$4.25
Blended $/1M (3∶1)
-
$2
Speed
tok/s
-
95
TTFT (s)
-
0.38
Reasoning
MMLU-Pro
—
—
GPQA Diamond
83.5
90.4
Humanity's Last Exam
22.0
45.5
AIME 2025
—
—
MATH-500
—
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
27.0
45.4
AA-LCR v1.1
83.3
79.0
CritPt
2.6
17.7
MMMU-Pro
74.3
—
IFBench
—
—
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
76.0
—
SWE-bench Pro
51.2
—
SWE-bench Multilingual
—
—
LiveCodeBench
—
—
Terminal-Bench 2.1
51.7
80.1
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
44.9
57.4
CursorBench
—
—
SWE-Rebench
—
—
NL2Repo-Bench
—
—
DeepSWE
—
55.0
WebDev Arena
1,360
1,535
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
1,427
1,500
| Metric | Muse Glimmer | Muse Spark 1.2 | Delta |
|---|---|---|---|
| Identity | |||
| Organization | Meta | Meta | — |
| Family | Llama / Muse | Llama / Muse | — |
| License | Apache 2.0 | Proprietary | — |
| Open weights | Yes | No | — |
| Release | Aug 10, 2026 | Aug 5, 2026 | — |
| Knowledge cutoff | 2026-01-04 | - | — |
| API / provider | No primary API price listed | Meta | — |
| Modalities | text, image → text | text, image, video, audio, pdf → text | — |
| Specs | |||
| Context window | 131K | 1.0M | B +918K |
| Max output | - | 131K | — |
| Parameters | 30B | — | — |
| Pricing | |||
| Input $/1M | — | $1.25 | — |
| Output $/1M | — | $4.25 | — |
| Blended $/1M (3∶1) | - | $2 | — |
| Speed | |||
| tok/s | - | 95 | — |
| TTFT (s) | - | 0.38 | — |
| Reasoning | |||
| MMLU-Pro | — | — | — |
| GPQA Diamond | 83.5 | 90.4 | B +6.9 pts |
| Humanity's Last Exam | 22.0 | 45.5 | B +23.5 pts |
| AIME 2025 | — | — | — |
| MATH-500 | — | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | 27.0 | 45.4 | B +18.4 pts |
| AA-LCR v1.1 | 83.3 | 79.0 | A +4.3 pts |
| CritPt | 2.6 | 17.7 | B +15.1 pts |
| MMMU-Pro | 74.3 | — | — |
| IFBench | — | — | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | 76.0 | — | — |
| SWE-bench Pro | 51.2 | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | — | — | — |
| Terminal-Bench 2.1 | 51.7 | 80.1 | B +28.4 pts |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | 44.9 | 57.4 | B +12.5 pts |
| CursorBench | — | — | — |
| SWE-Rebench | — | — | — |
| NL2Repo-Bench | — | — | — |
| DeepSWE | — | 55.0 | — |
| WebDev Arena | 1,360 | 1,535 | B +175 Elo |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | 1,427 | 1,500 | B +72 Elo |
Muse Glimmer
- Context
- 131K
- Parameters
- 30B
- Price
- —
- Speed
- —
- Modalities
- text, image → text
- License
- Apache 2.0
Meta's first open-weight model from Superintelligence Labs — 30B dense multimodal agentic model distilled from Muse Spark; Apache 2.0 with 128k context, built to run local agents on a single consumer GPU.
Muse Spark 1.2
- Context
- 1.0M / 131K out
- Parameters
- —
- Price
- $1.25 / $4.25
- Speed
- 95 tok/s · 0.38s TTFT
- Modalities
- text, image, video, audio, pdf → text
- License
- Proprietary
Co-trained with Muse Code for higher first-attempt accuracy and more reliable tool calling — 1M context coding-focused upgrade; Meta Model API at $1.25/$4.25 (contributor tier $0.10/$0.20).
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.