Model WarsJuly 8, 2026via MarkTechPost
SpaceXAI Releases Grok 4.5, a Cursor-Trained Model for Coding, Agentic Tasks, and Knowledge Work at $2/M Input
Why it matters
A new entrant in the coding/agentic model space with competitive pricing and specialized legal reasoning benchmarks signals intensifying competition in the agent-capable model tier. Cursor-trained focus suggests XAI is targeting developer workflows directly.
Key signals
- Model name: Grok 4.5
- Throughput: 80 TPS
- Pricing: $2 input / $6 output per million tokens
- Top ranking: #1 on Harvey's Legal Agent Benchmark
- Training approach: Cursor-trained
- Use cases: coding, agentic tasks, knowledge work
- Publisher: MarkTechPost (July 8, 2026)
The hook
SpaceXAI's Grok 4.5 just hit #1 on Harvey's Legal Agent Benchmark—and it's priced at $2/M input tokens.
SpaceXAI released Grok 4.5, a Cursor-trained model for coding, agentic tasks, and knowledge work. It serves at 80 TPS, costs $2/$6 per million tokens, and ranks #1 on Harvey's Legal Agent Benchmark.