Vorp Labs' release review confirms Muse Spark 1.2 shipped August 5, 2026, from Meta Superintelligence Labs as a coding-focused update to the July 9 Muse Spark 1.1, launched alongside the Muse Code beta terminal agent that is co-trained with the model. Standard-tier API pricing is unchanged from 1.1 at $1.25 input, $0.1 Meta's own launch charts place 1.2 second behind Claude Opus 5 on Terminal-Bench 2.1, DeepSWE v1.1, and Meta's internal coding bench, though Vorp flags those rankings as entirely vendor-run. The OpenRouter catalog lists meta/muse-spark-1.2 at the standard rates and context, while Meta has also released the separate Mus

Meta released Muse Spark 1.2 on August 5, 2026, alongside Muse Code, a beta terminal coding agent co-trained with the model. According to the launch guide, Meta published a 1,048,576-token (1M) context window and two API price tiers on the Meta Model API: a Standard tier at $1.25 per million input tokens and $4.25 per The guide notes that Meta's published benchmark charts arrived only as images in the launch materials, limiting reproducibility of coding and agentic scores. It walks through developer reaction in the first hours after launch and provides evaluation guidance for teams considering migration. The piece also references th

Benchable's catalog entry for Meta's Muse Spark 1.2 (endpoint meta/muse-spark-1.2-20260805, released August 5, 2026) describes it as a reasoning model designed for complex agentic tasks with a 1M-token context that accepts text, images, video, audio, and PDF inputs and returns text. Pricing is $1.25 prompt and $4.25 co The page reports 100% reliability across benchmarks and a 55th-percentile response time. Accuracy highlights include perfect General Knowledge, 97.0% Coding, 91.0% Instruction Following, 94.0% Reasoning, 93.0% Mathematics, and 98.0% Email Classification, though Email Classification only ranks at the 50th percentile. Pr

Artificial Analysis scored Muse Spark 1.2 at 54 on its Intelligence Index, a 3-point gain over Muse Spark 1.1 (51) and an 11-point gain over Muse Spark 1.0 (43, April 2026), placing Meta effectively tied with SpaceXAI for third among US labs and narrowly behind Claude Opus 5 (61), Claude Fable 5 (60), GPT-5.6 Sol (59), At Meta's $1.25/$4.25 per million token pricing, Muse Spark 1.2 costs $0.40 per Intelligence Index task — cheaper than GPT-5.5 ($1.18), Kimi K3 ($0.86), and GPT-5.6 Terra ($0.51), with only Grok 4.5 ($0.37) and GPT-5.6 Sol ($0.39) cheaper in its intelligence cluster; the increase over 1.1's $0.29/task is driven by 53%

The BenchLM catalog entry for Muse Spark 1.2 records a 1M-token context window, $1.25 input / $4.25 output per million tokens with $0.15 cached input and a $2.75 blended rate, 154 tokens per second throughput, and a 28.37-second time-to-first-token. Capability composite is 70.4/100 against a field median of 56.2, placi The page explicitly flags Muse Spark 1.2 as superseded by Muse Spark 1.3, signaling that newer Meta checkpoints exist in the family. Knowledge ranks as the model's strongest published category at the 10th position, while coding is its lowest eligible ranked category at 21. Of the benchmark rows tracked, eight are publi