Grok 4.5
Last verified August 15, 2026 · xAI pricing ↗$2.00/1M input · $6.00/1M output · 500K context · xAI
Count Tokens
—
Tokens
—
Input Cost
—
Output Cost
Estimate Monthly Cost
Monthly Cost Estimator
Quick:
<$0.0001/mo
Pick a preset above or enter custom usage
Alternatives to Grok 4.5
Pricing Details
xAIxAI's first model built specifically for coding and agent work, launched July 8, 2026 jointly with Cursor. Cached input $0.30/1M. Re-checked 2026-08-15 after Grok 4.6 shipped: unchanged and not retired. Still on xAI's live pricing table at these exact rates, still carries the same 150 RPS / 50M TPM limits as Grok 4.6, and still holds both the grok-4.5-latest and grok-build-latest aliases, so anything pinned to grok-build-latest is served this model and its cheaper $0.30 cache. The model page and rate card survive intact. xAI's sole retirement notice is dated 2026-05-15 and does not name it. Context window 500K (down from Grok 4.3's 1M). Above 200K prompt tokens the rate doubles to $4.00/$12.00 (cached $0.60), now published in xAI's own pricing table rather than inferred. Server-side tools: web/X search ~$5 per 1,000 calls. Artificial Analysis Intelligence Index 54 (#4). SWE-bench Pro 64.7, Terminal-Bench 2.1 83.3. Terminal-Bench 3.0 (official Harbor board, read 2026-08-17): 15.7% +/- 1.5 on the Cursor CLI harness, a 67.6-point drop from the same model's 2.1 score and the clearest evidence that the two versions are not comparable. Its $766.02 run cost is the lowest on that board but the benchmark authors exclude the row from their own Pareto chart over a cost reporting issue in the Cursor CLI harness, so treat it as unusable. ~4x more token-efficient than Opus 4.8 per AA. Trained on Cursor session data (Cursor disclosed accidental codebase-snapshot contamination). Estimated tokens (different tokenizer).
Input / 1M tokens
$2
Output / 1M tokens
$6
Context Window
500K
Max Output
16K
Price History
No price change recorded since 2026-07-27, when this site started tracking this model; it launched on 2026-07-08.
Frequently Asked Questions
Common questions about Grok 4.5 pricing and usage
Read More About Grok 4.5
Terminal-Bench 3.0 publishes what every run cost, which almost no other leaderboard does. Two of its ten rows are billed at a card OpenAI retired on July 30, and repricing them makes the ninth-place model the cheapest per solved task by 3.5x.
August 17, 2026 · 11 min read
Grok 4.6 costs exactly what Grok 4.5 costs on input, on output, and at the cliff. The only line that moved is cached input, up 67%, and that is the line an agent loop actually lives on.
August 15, 2026 · 11 min read
Claude Opus 4.8 just slipped off the LLM value frontier. A $3 model scores higher on the intelligence index and costs a quarter less to run, and it is not the only flagship the math has passed.
July 23, 2026 · 9 min read