4 September 2026
Meta releases Muse Spark 1.3, claims parity with top AI models
First reported
InfoWorld and SiliconANGLE ran this on , a day before the other 2 sources picked it up.
- Meta released Muse Spark 1.3, its most powerful model yet, claiming performance matching Anthropic's Claude and surpassing OpenAI's latest in coding tasks.
- The model uses about 25% fewer tokens than its predecessor to complete the same work, potentially lowering costs for developers at unchanged pricing.
- Independent benchmarking ranked Muse Spark 1.3 third overall after Claude Fable 5.1 and Claude Opus 5, ahead of OpenAI's models.
- A researcher recently demonstrated that similar models remain vulnerable to jailbreak attacks, though Meta's latest versions resisted the exploit tested.
Where they differ
TLDR AIhighlighted security vulnerabilities in frontier models including Meta's earlier Spark 1.1.
AI Breakfastfocused on the competitive capabilities and pricing of Spark 1.3 itself. Both facts are true and complementary, not contradictory.
What each one reported
A MATS researcher found that a synthetic transcript generation prompt could be turned into a universal jailbreak template hitting 84-100% attack success on nine of 23 vulnerable models tested. Only recent Anthropic models and Meta Muse Spark 1.1 were never fully broken.
Meta released Muse Spark 1.3 with competitive Intelligence Index scores matching GPT-5.6 Sol and Claude Opus 5, priced aggressively at $1.25 and $4.25 per million tokens. The company also holds a higher-capability max variant pending safety testing and tied the release to personal agent roadmaps.
Reported by SiliconANGLE, InfoWorld