Breaking

OpenClaw v2026.7.1 Adds Support for Meta's Muse Spark 1.1 Agentic Reasoning Model

July 14, 2026 at 21:32 EDT

AI Agents
Foundation Models
Open Source

OpenClaw's latest release, v2026.7.1, now natively supports Meta's Muse Spark 1.1, bringing the company's new multimodal reasoning model to the open-source agent framework for coding, tool use, and computer-use workflows.

v2026.7.1 · OpenClaw

OpenClaw goes native with Meta's Muse Spark 1.1

The privacy-first, messaging-native open-source agent framework now runs Meta's agent-tuned multimodal model directly — with streaming, tool calls, images, and reasoning controls for coding, tool use, and computer-use workflows.

token context window, up sharply from v1.0

Intelligence Index, up 8 points from v1.0

532

contributors · 3,063 contributions

Benchmark profile

Strong on agentic tool use — but not the top on pure coding.

88.1

MCP Atlas
scaled tool use

54.7

JobBench
exceeds Opus 4.8 in cases

Intelligence Index
on par w/ GLM-5.2, GPT-5.4

Meta Model API pricing · per 1M tokens

$1.25

input tokens

$4.25

output tokens

$20

free credits for new users

Cache-hit discounts included · widely seen as strong on cost efficiency and token usage.

Praised for

Fast, powerful, well-suited to running agents
High-quality tool use at low cost
Persistent memory & 24/7 automation
1M context, parallel sub-agents, native computer use

Where it trails

Not the top performer on pure coding tasks
Opus 4.8 & GPT-5.5 remain stronger there
Behind Grok 4.5 on the Intelligence Index
Earlier setup complexity & immature interface

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.