AI

Meta's 816B Muse Spark Open scores 39 on Artificial Analysis, the 180B Qwen3.8-Flash-Next scores 40. Score vs size for 11 open-weight models

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 23/10/2026, 15:47:25 UTC.

Briefings

Meta's Artificial Analysis ranks open-weight models by total parameters, with Muse Spark Open (816B) scoring 39 and Qwen3.8-Flash-Next (180B) scoring 40. The chart uses an aggregate evaluation metric from AA's own runs, showing 11 models released in October 2026. Key facts include memory usage for Unsloth quantized versions and Meta's recommended hardware setup for Muse Spark Open.

Key points

• Muse Spark Open (816B total, 42B active) scored 39 on Artificial Analysis, while Qwen3.8-Flash-Next (180B total, 6B active) scored 40.
• Unsloth's 4-bit quantized versions require 16-19 GB of memory, and Meta recommends two M5 Ultra Mac Studios with 512 GB each for Muse Spark Open.

Context

This chart compares open-weight models based on total parameters using an aggregate evaluation metric from Artificial Analysis's own runs, released October 2026. It highlights performance differences between large models like Muse Spark Open and Qwen3.8-Flash-Next, along with memory requirements for quantized versions and recommended hardware for specific models.

Limitations

The evaluation metric is an aggregate number from AA's own runs and does not account for speed, long context, vision capabilities, or specific task performance. Memory usage varies significantly between quantized versions, and the chart only covers 11 models released in October 2026.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 23/10/2026, 16:17:08 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeifjzajhxehco3ljp54oqmbt7fefrdabehyh7fmxasugcqnyay6tam