What the vendor claims, next to what independent boards measured. The two are never mixed in one column.
Source: OpenRouter, as of Oct 28, 09:19 UTC. Model page on OpenRouter
Quoted word for word from the vendor's own launch post or model card. Claims, not measurements.
No vendor claims recorded yet. They come from launch posts and model cards the Drops covered.
Measured by independent sources (Epoch AI, LMArena, OpenRouter's benchmark runs, τ²-bench, Vectara), not by the vendor. "Listed as" is the source's own name for the model, so you can check our match.
| Board | Result | Rank |
|---|---|---|
| OpenRouter · GPQA DiamondListed as: Muse Spark Openvia Modal, the most accurate providerSource: OpenRouter (openrouter.ai), licensed under CC BY 4.0, as of Oct 28, 2026. Source | 90.6%Accuracy | 35 |
| OpenRouter · GPQA Diamond: cost per taskListed as: Muse Spark Openvia Modal, the most accurate providerLower is better on this board.Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0, as of Oct 28, 2026. Source | $0.066Cost per task | 34 |
| OpenRouter · GPQA Diamond: time per taskListed as: Muse Spark Openvia Modal, the most accurate providerLower is better on this board.Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0, as of Oct 28, 2026. Source | 65 sTime per task | 21 |
| OpenRouter · τ²-bench airlineListed as: Muse Spark Openvia Fireworks, the most accurate providerSource: OpenRouter (openrouter.ai), licensed under CC BY 4.0, as of Oct 28, 2026. Source | 77.0%Accuracy | 21 |
| OpenRouter · τ²-bench airline: cost per taskListed as: Muse Spark Openvia Fireworks, the most accurate providerLower is better on this board.Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0, as of Oct 28, 2026. Source | $0.10Cost per task | 28 |
| OpenRouter · τ²-bench airline: time per taskListed as: Muse Spark Openvia Fireworks, the most accurate providerLower is better on this board.Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0, as of Oct 28, 2026. Source | 175 sTime per task | 22 |
OpenRouter's list price per 1M tokens in each of our snapshots.
Unchanged across our 25 snapshots.
| Snapshot | Input $/1M | Output $/1M |
|---|---|---|
| Oct 28, 09:22 UTC | $1.25 | $4.25 |
| Oct 27, 16:03 UTC | $1.25 | $4.25 |
| Oct 27, 04:03 UTC | $1.25 | $4.25 |
| Oct 26, 16:04 UTC | $1.25 | $4.25 |
| Oct 26, 04:03 UTC | $1.25 | $4.25 |
| Oct 25, 16:04 UTC | $1.25 | $4.25 |
| Oct 25, 06:03 UTC | $1.25 | $4.25 |
| Oct 24, 15:04 UTC | $1.25 | $4.25 |
| Oct 24, 03:03 UTC | $1.25 | $4.25 |
| Oct 23, 15:03 UTC | $1.25 | $4.25 |
| Oct 23, 03:04 UTC | $1.25 | $4.25 |
| Oct 22, 15:03 UTC | $1.25 | $4.25 |
| Oct 22, 05:04 UTC | $1.25 | $4.25 |
| Oct 21, 15:03 UTC | $1.25 | $4.25 |
| Oct 21, 03:02 UTC | $1.25 | $4.25 |
| Oct 20, 15:04 UTC | $1.25 | $4.25 |
| Oct 20, 03:03 UTC | $1.25 | $4.25 |
| Oct 19, 15:04 UTC | $1.25 | $4.25 |
| Oct 19, 10:50 UTC | $1.25 | $4.25 |
| Oct 18, 15:03 UTC | $1.25 | $4.25 |
| Oct 18, 03:05 UTC | $1.25 | $4.25 |
| Oct 17, 15:03 UTC | $1.25 | $4.25 |
| Oct 17, 03:03 UTC | $1.25 | $4.25 |
| Oct 16, 15:02 UTC | $1.25 | $4.25 |
| Oct 16, 03:03 UTC | $1.25 | $4.25 |
Source: OpenRouter, as of Oct 28, 09:22 UTC
Apple's ML research team says four M5 Ultra Mac Studios run Meta's 816B model at over 36 tokens per second with MLX. AWS added it to SageMaker JumpStart. In Epoch AI's run it scored 0.886 on GPQA diamond.
Why it matters: A frontier-class open model can now run on desk hardware or in your own cloud account, keeping code private. CNBC says the required Mac Studio starts at $10,799.
From Drop #028 · Fri, Oct 16 · Evening