Meta

Meta’s largest open-weights instruct checkpoint in the Llama 3.1 family, aimed at strong reasoning and coding quality with a permissive license for research and customization. It is typically served on dedicated GPU clusters or via partners (cloud inference, on-prem) rather than a single vendor API.

Llama 3.1 8B Instruct is a small open-weights model for edge laptops, single-GPU servers, and ultra-low-latency assistants. Quality per dollar is competitive for simple tasks but not for frontier reasoning.

Llama 3.1 70B Instruct is a mid-size open-weights instruct model balancing quality and deployability on a single large GPU or small multi-GPU nodes. Common for private assistants, on-prem pilots, and fine-tunes where 405B is impractical.

Muse Glimmer 30B is Meta Superintelligence Labs' open-weight multimodal model for local agents, coding, tool use, long-horizon reasoning, and image understanding. Meta's model card documents a dense 29.6B-parameter architecture with a dedicated perception encoder, 131,072+ context, text-and-image input, text output, controllable reasoning effort, and Apache 2.0 weights.

Muse Spark 1.3 is Meta's September 2, 2026 hosted model for long-horizon agentic and coding work. The API ID is muse-spark-1.3, with a 1,048,576-token context window, text/image/video/PDF input, and text output. Meta documents it in Muse Code and Meta Model API, with reasoning effort including max on the Standard tier. Official docs still list muse-spark-1.2 as the Muse Code default while recommending 1.3 for new API work.

API ID muse-spark-1.3

Context tokens

1,048,576

Muse Image 1.0 is Meta's hosted image generation and editing model on Meta Model API. The API ID is muse-image-1.0. One model handles text-to-image, image editing, and multi-image composition. Meta documents an agentic loop that can search the web for visual references and run code to plan layouts before rendering. Reach it through the Responses API or POST /v1/images/generations and /v1/images/edits.

API ID muse-image-1.0

Muse Voice Transcribe is Meta Superintelligence Labs' September 1, 2026 real-time audio perception model. The API ID is muse-voice-transcribe-1.0. It does streaming ASR with adaptive delay, speaker diarization for 20+ speakers, and native endpointing. Access is via Meta Model API (WebSocket realtime and file POST /v1/asr/transcribe), Meta AI for Mac, and Muse Code dictation.

API ID muse-voice-transcribe-1.0