xAI ships Grok 4.5 for coding, agentic tasks, and knowledge work

xAI has made Grok 4.5 generally available on the xAI API, positioning the model as its new option for coding, agentic tasks, and knowledge work, according to the company's own release notes.

What's new

xAI's release notes state plainly: "Grok 4.5, SpaceXAI's model for coding, agentic tasks, and knowledge work, is now available on the xAI API." (SpaceXAI is the company's current corporate branding following its consolidation with Musk's other ventures.) Alongside general availability, xAI published:

• Pricing: $2 per 1 million input tokens, $6 per 1 million output tokens
• Configurable reasoning effort: low, medium, or high, with high set as the default
• Documentation: a dedicated Grok 4.5 developer overview page and a separate consumer-facing announcement

The model is live on the API today; xAI's release notes place the update under its July changelog entries, alongside continued iteration on the Grok family.

Context

Grok 4.5 follows Grok 4.3, which reached Amazon Bedrock with a 1-million-token context window and configurable reasoning earlier this year. That release already signaled xAI's push to compete on both context length and reasoning flexibility rather than a single flagship benchmark number. Grok 4.5 extends that same reasoning-effort dial (low/medium/high) into what xAI now frames as its coding- and agent-focused model, continuing a cadence of frequent point releases rather than infrequent, monolithic launches — a pattern also visible in xAI's recent Grok Voice and Grok Build releases.

The configurable reasoning-effort mechanic mirrors an industry-wide shift: OpenAI, Anthropic, and Google have all shipped similar reasoning-budget controls on their frontier models over the past year, letting developers trade latency and cost against answer quality on a per-request basis. xAI's decision to expose the same three-tier control on Grok 4.5, paired with per-token pricing that undercuts several rival reasoning-tier models, suggests the company is competing primarily on cost-to-capability ratio for developer workloads rather than chasing a single leaderboard score.

Why it matters

A GA release aimed explicitly at "coding, agentic tasks, and knowledge work" puts Grok 4.5 in direct contention with the coding- and agent-oriented tiers of GPT-5-class and Claude models, where developer adoption — not just chatbot usage — drives revenue. Configurable reasoning effort gives builders a lever to control cost on high-volume agentic pipelines, which is increasingly a deciding factor for teams choosing an API provider at scale.

The pricing xAI has set — $2/$6 per million tokens — lands in a competitive band for a reasoning-capable model, and shipping it as a straightforward API update (rather than a splashy standalone event) fits the faster release cadence xAI has adopted through 2026. For developers already building agentic tooling on Grok, this is a low-friction upgrade path; for the broader market, it's another data point in the ongoing price and reasoning-flexibility race among the four frontier labs.