Grok 4.5 — xAI moves upmarket but stays the cheap frontier option

xAI shipped Grok 4.5 on July 8 — its new flagship for coding, agentic tasks, and knowledge work, priced at $2 per 1M input / $6 per 1M output with a 500K context window. That is a real price step up from Grok 4.3 ($1.25/$2.50), but xAI's pitch is that roughly 2× token efficiency — solving tasks in under half the steps — makes the effective cost competitive anyway. API-only, no open weights (xAI has open-sourced nothing since Grok-1 in 2024), so no local pick moves; the cost calculator row moves from 4.3 to 4.5.

Verdict: xAI's new flagship at $2/$6 with a 2× token-efficiency claim — hosted-only, priced between the budget and mid API tiers

The take

The facts, verified against x.ai/news/grok-4-5 (July 8, 2026): $2/$6 per 1M, 500K context, configurable reasoning effort (low/medium/high), model id `grok-4.5`, available day one in the xAI console/API, Grok Build, and Cursor. Not yet available in the EU at launch (xAI says mid-July). On xAI's posted DeepSWE 1.0 numbers it lands at 62.0% — behind Claude Fable 5 (66.1%) and GPT-5.5 at max effort (64.3%) but ahead of Opus 4.8 (55.8%); the efficiency claim, not the raw score, is the sales pitch.

The pricing position is the story: at a $4 blended average per 1M it sits between the budget tiers (GPT-5.6 Luna $3.50, Haiku 4.5 $3) and the mid tiers (Gemini 3.5 Flash $5.25, Sonnet 5 intro $6) while claiming frontier-class capability. Grok 4.3 remains served at $1.25/$2.50 for now, so the cheapest-frontier crown xAI held isn't gone — it has just split into "cheapest" (4.3, superseded) and "current" (4.5).

Our call: no local-pick change; the cost calculator's xAI row is now Grok 4.5 at $2/$6. Same honest framing as every hosted drop this month: a cheaper, more efficient cloud tier stretches the payback on buying hardware if your usage is light — and does nothing to the case for local if your reasons are privacy, offline reliability, or high-volume agent loops where token bills compound.

Where this fits

Models: GLM-5.1 · Kimi K2.6 · Qwen3-Coder-30B-A3B · North Mini Code (30B-A3B)

Hardware: NVIDIA RTX 5090 · NVIDIA RTX 4090 · Mac Studio M4 Max 64 GB

Sources

Next step

Try this in the planner→