# Grok 4.5

SpaceXAI, released 2026-07-08

**Verdict: conditional.** Competitively priced coding and agentic model at $2/$6 per 1M tokens up to 200K tokens of context, doubling to $4/$12 above that, with a 500K window. xAI's published benchmarks place it mid-pack behind Fable 5 and GPT-5.5, though it leads Opus 4.8 on DeepSWE and Terminal-Bench 2.1. The real story is cost efficiency: 4.2x more token-efficient than Opus 4.8 on SWE-Bench Pro. Best for teams needing strong coding at a fraction of frontier pricing; xAI's docs list only US regions. Last verified 2026-07-09.

## Pricing

- Input: $2 / 1M tokens
- Output: $6 / 1M tokens
- Context: 500K ctx; 200K+ $4/$12; cache $0.30 (200K+ $0.60)

## Strengths

- Cost-sensitive coding workloads — $2/$6 pricing undercuts Opus by 60% on input, 76% on output
- Token-efficient agentic tasks — resolves SWE Bench Pro with ~16K output tokens, 4.2x fewer than Opus 4.8
- Long-running tool-use and multi-step engineering — trained alongside Cursor for agentic coding workflows
- Office productivity — built-in Excel, PowerPoint, Word integration via Grok Build
- Prompts of 200K tokens or more budgeted off the headline rate: input and output double to $4 and $12 per 1M, and cached input rises from $0.30 to $0.60
- Teams needing the absolute frontier — Fable 5 and GPT-5.5 lead on most coding benchmarks
- EU-based projects — xAI's grok-4.5 docs list only the us-east-1 and us-west-2 regions
- Workloads requiring independent benchmark verification — all current scores are vendor-reported

## Benchmarks

- DeepSWE 1.0: 62.0% (https://x.ai/blog/grok-4-5)
- DeepSWE 1.1: 53% (https://x.ai/blog/grok-4-5)
- SWE Marathon: 29.0% (https://x.ai/blog/grok-4-5)
- Terminal Bench 2.1: 83.3% (https://x.ai/blog/grok-4-5)
- SWE Bench Pro: 64.7% (https://x.ai/blog/grok-4-5)

## Sources

- [SpaceXAI Docs — grok-4-5 model page](https://docs.x.ai/developers/grok-4-5)
- [Roo (beehiiv) — Grok 4.5 benchmark deep-dive](https://roo.beehiiv.com/p/grok-4-5)
- [Digital Applied — Grok 4.5 spec sheet (context window, pricing, capabilities)](https://www.digitalapplied.com/blog/grok-4-5-spacexai-coding-office-agent-launch-2026)
- [TechCrunch — SpaceXAI releases Grok 4.5, which Elon describes as an 'Opus-class model'](https://techcrunch.com/2026/07/08/spacexai-releases-grok-4-5-which-elon-describes-as-an-opus-class-model/)
- [The Next Web — SpaceXAI launches Grok 4.5](https://thenextweb.com/news/spacexai-grok-4-5-cursor-opus-class-coding-model)
- [Axios — Scoop: Musk's SpaceXAI releases new model, Grok 4.5](https://www.axios.com/2026/07/08/spacexai-grok-new-model)
- [xAI docs: Grok 4.5 model](https://docs.x.ai/docs/models/grok-4.5)
- [LM Market Cap — Grok 4.5 pricing and benchmarks](https://lmmarketcap.com/model/grok-4-5)
- [xAI models and pricing](https://docs.x.ai/docs/models)
- [SpaceXAI — Introducing Grok 4.5 (official blog)](https://x.ai/blog/grok-4-5)
- [Artificial Analysis — Grok 4.5 intelligence/performance metrics](https://artificialanalysis.ai/models/grok-4-5)

Canonical: https://neomanex.com/models/grok-4-5