← All models

Grok 3

Language Retired Full comparison ↗

xai/grok-3 · by xAI · decoder-only-transformer

Pricing — 1 offering(s)

xAI

Input tokens

Retired — no longer sold · last price $3.00 / 1M tokens (input), until 2026-05-14

  • $3.00 / 1M tokens (input) Historical 2025-02-17 → 2026-05-14

Output tokens

Retired — no longer sold · last price $15.00 / 1M tokens (output), until 2026-05-14

  • $15.00 / 1M tokens (output) Historical 2025-02-17 → 2026-05-14

Showing the active price and any recorded history. Full pricing history is available via the paid API — see API docs.

See how Grok 3 fits into a cost-aware routing setup

See how →

Capability profile

How Grok 3 rates across core capability dimensions, with the task-level evidence behind each rating.

reasoning Strong
coding Moderate

Aider Polyglot (225 Exercism exercises across 12 languages; the model gets a spec and hidden failing tests and must write a passing implementation): resolves 53.3% — fails roughly 1 exercise in 2. Independent (Aider leaderboard, "Grok 3 Beta", 2025-04). No SWE-bench score of any kind has been published for Grok 3.

tool use Strong

xAI API explicitly documents function calling for Grok-3 (x.ai/api). Upgraded from moderate: named in API docs, not just a generic capability mention.

instruction following Strong
context window Moderate

131K context.

multilingual Moderate
speed Moderate
cost efficiency Moderate

Coding benchmarks

View the full coding benchmark leaderboard →
Benchmark Score Harness Source
livecodebench ↗ 57.0% 2025-02 — 📊 leaderboard ↗
aider-polyglot ↗ 53.3% 2025-04 aider 📊 leaderboard ↗
bigcodebench ↗ 27.0% 2025-04 — 📊 leaderboard ↗

Operator guidance

Competitive with Claude Sonnet and GPT-4o at the same price tier. Unique advantage is real-time X training data. Route here when current events context matters or for diversification across frontier providers.

Use cases

Limitations

Citations