Days after Meta released Muse Spark Open — an 816B-total/42B-active-parameter model with a 1M-token context window that Meta calls the most capable US open-weight model to date — attention is shifting from raw performance to the fine print.

According to Interconnects, the Muse Model License is 'pretty much all of the restrictive license behaviors we've seen' in recent open-model releases, and Nathan Lambert argues it may be the one thing that holds the model back.

According to Interconnects, the license requires products built on the model to display 'Built with Muse' branding — a rule Kimi tried and almost no one complies with. It gates commercial use: companies earning more than $20M per month across their entire business need a separate license with Meta.

An acceptable use policy makes the model 'extremely not open-source' by OSS definitions, and the restrictions extend to outputs and derivatives, creating legal ambiguity over synthetic data generation. Internal research and applications remain permitted.

Interconnects frames Muse Spark Open as a step change: before it and Reflection's Beam, there were 20+ open-weight models from Chinese companies better than the best Western ones, and the gap between the best American and Chinese models has been reduced roughly by half.

Per tech.co, the model scored 39 on Artificial Analysis's Intelligence Index — ahead of several Western open-weight models, but still behind z.ai's GLM 5.5 and Moonshot AI's Kimi K3. Interconnects places its performance above GLM 5.2, near DeepSeek V4.1 Flash, and behind GLM 5.3 or Kimi K3.

The model card, cited by Interconnects, says this open version received additional safety post-training to resist misuse as a standalone model without system guardrails, and Meta recommends deploying it within a broader AI system.

Observers speculated the release was a model 'broken' during safety training, but Interconnects argues it was more likely a safety tune of Muse Spark 1.2 costing only about one point on the Artificial Analysis index.

Meta says it was evaluated under its Superintelligence Scaling Framework, per t2ONLINE.

Interconnects calls the license 'complex and unproven in court' and predicts many players will likely ignore it, building on the model through inference providers while Meta pursues only egregious violations.

Historically, licenses have not determined AI adoption, but the publication argues non-standard licenses are an unnecessary strategic risk for a lab trying to regain adoption — and warns that could change in the future.

Muse Spark Open is Meta's second open-weight release since its AI overhaul, after the smaller Muse Glimmer in August, and the company is now the only American lab with frontier open and closed models.

Interconnects ranks the field as Anthropic > OpenAI > Meta ~ Gemini > xAI ~ Chinese Open Weights > U.S. Open Weights, and reports rumors of more strong American open-weight models incoming — including Mistral Large 4, announced with an index score of 38 but with weights not yet released.