Grok 4.7 vs Muse Glimmer 30B
Compare Grok 4.7 and Muse Glimmer 30B side-by-side.
Compare Grok 4.7 vs Muse Glimmer 30B live
Run the same image across every model that supports a task and compare their outputs side-by-side.
These models don't share enough common tasks for a side-by-side demo. See the comparison table below for their capabilities.
Models in this comparison
Grok 4.7 vs Muse Glimmer 30B on Vision Evals
Grok 4.7 scores higher on 3 of the six Vision Evals tasks.
The widest gap is Reasoning, where Grok 4.7 leads 64.2% to 57.6%.
Overall, Grok 4.7 averages 71.9% (#22 of 54) against 70.8% (#24 of 54) for Muse Glimmer 30B.
Muse Glimmer 30B is both cheaper ($0.0011 vs $0.012 per sample) and faster (8.7s vs 23.6s per sample).
Grok 4.7 vs Muse Glimmer 30B Comparison Table
Evals updated September 22, 2026Pricing updated September 22, 2026
| Property | Grok 4.7 | Muse Glimmer 30B |
|---|---|---|
| Organization | SpaceXAI | Meta |
| Category | closed | open |
| Modality | — | multimodal |
| Release Date | Sep 2026 | Aug 2026 |
| Context Window | 500K | 131K |
| Parameters | 29.6B | |
| License | Proprietary | Apache 2.0 |
| Pricing per 1M tokens | ||
| Input $/1M | $1.60 | $0.300 |
| Output $/1M | $4.80 | $1.20 |
| Vision Tasks | ||
| Captioning | Demo | |
| Chart Question Answering | ||
| Classification | Demo | |
| Document Question Answering | ||
| Image Tagging | ||
| Multi-Label Classification | ||
| Object Detection | Demo | |
| OCR | Demo | |
| Vision Language | ||
| Visual Question Answering | Demo | |
| Model Features | ||
| Foundation Vision | ||
| LLMs with Vision Capabilities | ||
| Multimodal Vision | ||
Vision Evalsground-truth scores across 6 vision tasks, pooled at low effort | ||
| Overall | 71.9% | 70.8% |
| Avg cost / sample | $0.012 | $0.0011 |
| Avg speed / sample | 23.55s | 8.70s |
| By task | ||
| Object Detection (low) | 40.4% ±0.6, Mean of 3 runs, range 39.8 to 41.0 | 41.0% |
| Object Detection (high) | 41.2% ±1.6, Mean of 3 runs, range 39.6 to 42.8 | – |
| Counting (low) | 61.7% ±1.3, Mean of 3 runs, range 60.8 to 63.5 | 66.2% |
| Counting (high) | 60.8% ±1.3, Mean of 3 runs, range 59.5 to 62.2 | – |
| Identification (low) | 87.5% ±3.1, Mean of 3 runs, range 84.4 to 90.6 | 81.3% |
| Identification (high) | 80.2% ±1.6, Mean of 3 runs, range 78.1 to 81.3 | – |
| OCR (low) | 92.6% ±0.7, Mean of 3 runs, range 92.1 to 93.4 | 92.1% |
| OCR (high) | 93.5% ±0.3, Mean of 3 runs, range 93.1 to 93.8 | – |
| Data Extraction (low) | 84.9% ±2.6, Mean of 3 runs, range 82.5 to 87.6 | 86.6% |
| Data Extraction (high) | 87.6% ±1.5, Mean of 3 runs, range 86.6 to 89.7 | – |
| Reasoning (low) | 64.2% ±2.3, Mean of 3 runs, range 62.3 to 66.9 | 57.6% |
| Reasoning (high) | 66.9% ±1.3, Mean of 3 runs, range 65.6 to 68.2 | 62.9% |
Grok 4.7 vs Muse Glimmer 30B: Overview
Grok 4.7 is a proprietary model from SpaceXAI, released on September 21, 2026. It accepts text and images as input and returns text. It extends Grok 4.6 and is listed at the same API price.
Its Vision Evals scores are on the leaderboard. Running it in the Playground is not available yet, because the inference workflow is not ready.
Muse Glimmer 30B is a dense vision language model from Meta built for long-horizon agentic work on local hardware. The architecture pairs a 52-layer causal text decoder with a roughly 1.8B parameter ViT-G/14 perception encoder for about 29.6 billion parameters in total, and it accepts interleaved text and image input so an agent can interpret screenshots, charts, and documents alongside conversation. The decoder uses grouped-query attention with 32 query heads and 2 key-value heads, a repeating pattern of three sliding-window local attention layers followed by one global layer, SwiGLU feed-forward blocks, and rotary position embeddings applied on the local layers, supporting a trained context of 131,072 tokens.
Meta describes the model as distilled from the larger Muse Spark and trained and evaluated around agentic behavior: end-to-end task completion, schema-accurate tool calling, multi-step reasoning across extended workflows, and recovery when a tool call returns an unexpected result. Reasoning effort is selectable across low, medium, high, and xhigh settings, and the model emits channel-scoped reasoning traces together with XML style tool calls rather than JSON, which requires parsers specific to this family. A companion block-diffusion drafter head predicts blocks of 16 tokens per forward pass for speculative decoding, with the main model verifying the proposals in parallel.
Frequently Asked Questions
On Roboflow's Vision Evals, Grok 4.7 performed slightly better overall. The two split the six vision tasks 3 to 3, but Grok 4.7 averages 71.9% (#22 of 54) against 70.8% (#24 of 54) for Muse Glimmer 30B. These scores measure vision capabilities only; either model may perform differently on work outside this benchmark's tasks.
Yes. On the Vision Evals Reasoning benchmark at low effort, Grok 4.7 leads with 64.2% against 57.6%. This is the widest gap between the two models across the benchmark's tasks.
Muse Glimmer 30B is cheaper to run. Across Roboflow's Vision Evals task mix it averaged $0.0011 per sample against $0.012. Grok 4.7 is priced at $1.60 per 1M input tokens and $4.80 per 1M output; Muse Glimmer 30B is priced at $0.30 per 1M input tokens and $1.20 per 1M output. Actual costs depend on your image sizes, prompts, and output length.
Muse Glimmer 30B is faster. Across Roboflow's Vision Evals it averaged 8.7s per inference against 23.6s. Latency varies with image size, prompt length, and provider load, so treat these as relative rather than guaranteed figures.