vs
0:4
overall capability wins (0 ties across 4 scored). Muse Spark 1.3 leads overall.
- Reasoning0:3
- Pricing3:0
- Specs0:1
Verdict
Muse Spark 1.3 edges reasoning & knowledge; Muse Spark 1.3 offers ~128.0x larger context (1.0M vs 8K); Llama 3 70B ships open weights (Llama 3).
- Muse Spark 1.3 edges reasoning & knowledge
- Muse Spark 1.3 offers ~128.0x larger context (1.0M vs 8K)
- Llama 3 70B ships open weights (Llama 3)
- Llama 3 70B is ~2.3x cheaper on a blended token basis than Muse Spark 1.3
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- Llama 3 70B
- $1.10
- Muse Spark 1.3
- $2.31
Llama 3 70B is about 2.1x cheaper on this workload.
Price per 1M tokens
Full comparison
Organization
Meta
Meta
Family
Llama / Muse
Llama / Muse
License
Llama 3
Proprietary
Open weights
Yes
No
Release
Apr 18, 2024
Sep 2, 2026
Knowledge cutoff
-
-
API / provider
Together / Fireworks (ref.)
Meta
Modalities
text → text
text, image, video, pdf → text
Specs
Context window
8K
1.0M
Max output
8K
-
Parameters
70B
—
Pricing
Input $/1M
$0.88
$1.25
Output $/1M
$0.88
$4.25
Blended $/1M (3∶1)
$0.88
$2
Speed
tok/s
55
-
TTFT (s)
0.4
-
Reasoning
MMLU-Pro
—
—
GPQA Diamond
37.9
93.5
Humanity's Last Exam
4.5
48.7
AIME 2025
—
—
MATH-500
—
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
17.7
43.6
AA-LCR v1.1
—
83.0
CritPt
—
24.9
MMMU-Pro
—
—
IFBench
37.1
—
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
—
—
SWE-bench Multilingual
—
—
LiveCodeBench
—
—
Terminal-Bench 2.1
—
85.8
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
43.6
—
SciCode
—
58.8
CursorBench
—
—
SWE-Rebench
—
—
NL2Repo-Bench
—
—
DeepSWE
—
—
WebDev Arena
—
—
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
1,276
—
| Metric | Llama 3 70B | Muse Spark 1.3 | Delta |
|---|---|---|---|
| Identity | |||
| Organization | Meta | Meta | — |
| Family | Llama / Muse | Llama / Muse | — |
| License | Llama 3 | Proprietary | — |
| Open weights | Yes | No | — |
| Release | Apr 18, 2024 | Sep 2, 2026 | — |
| Knowledge cutoff | - | - | — |
| API / provider | Together / Fireworks (ref.) | Meta | — |
| Modalities | text → text | text, image, video, pdf → text | — |
| Specs | |||
| Context window | 8K | 1.0M | B +1.0M |
| Max output | 8K | - | — |
| Parameters | 70B | — | — |
| Pricing | |||
| Input $/1M | $0.88 | $1.25 | A +$0.37 |
| Output $/1M | $0.88 | $4.25 | A +$3.37 |
| Blended $/1M (3∶1) | $0.88 | $2 | A +$1.12 |
| Speed | |||
| tok/s | 55 | - | — |
| TTFT (s) | 0.4 | - | — |
| Reasoning | |||
| MMLU-Pro | — | — | — |
| GPQA Diamond | 37.9 | 93.5 | B +55.6 pts |
| Humanity's Last Exam | 4.5 | 48.7 | B +44.2 pts |
| AIME 2025 | — | — | — |
| MATH-500 | — | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | 17.7 | 43.6 | B +25.9 pts |
| AA-LCR v1.1 | — | 83.0 | — |
| CritPt | — | 24.9 | — |
| MMMU-Pro | — | — | — |
| IFBench | 37.1 | — | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | — | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | — | — | — |
| Terminal-Bench 2.1 | — | 85.8 | — |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | 43.6 | — | — |
| SciCode | — | 58.8 | — |
| CursorBench | — | — | — |
| SWE-Rebench | — | — | — |
| NL2Repo-Bench | — | — | — |
| DeepSWE | — | — | — |
| WebDev Arena | — | — | — |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | 1,276 | — | — |
Llama 3 70B
- Context
- 8K / 8K out
- Parameters
- 70B
- Price
- $0.88 / $0.88
- Speed
- 55 tok/s · 0.4s TTFT
- Modalities
- text → text
- License
- Llama 3
Original Llama 3 70B instruct — widely fine-tuned open baseline before 3.1 / 3.3 / 4.
Muse Spark 1.3
- Context
- 1.0M
- Parameters
- —
- Price
- $1.25 / $4.25
- Speed
- —
- Modalities
- text, image, video, pdf → text
- License
- Proprietary
Meta's Sept 2026 coding/agentic upgrade to Muse Spark — ~20% fewer tool calls and ~25% fewer tokens than 1.2 in internal comparisons, with longer-horizon multi-workflow persistence and more proactive clarifying questions. Same 1M context and $1.25/$4.25 Meta Model API pricing as 1.2 (contributor tier $0.10/$0.20). Open weights planned but not yet released; max reasoning mode pending further safety testing.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.