Grok 4.3 is xAI's mid-tier reasoning workhorse, priced well under Claude Sonnet 5.5 on output. Its cache discount is shallower and its rates double on long prompts, which narrows the gap on some workloads.
Grok 4.3 is 4× cheaper on output and 1.6× cheaper on input than Claude Sonnet 5.5. Sonnet 5.5 has the cheaper cache hit ($0.10 vs $0.20) and no long-context surcharge. Once a prompt reaches 200K tokens Grok's rates double, which puts its input above Sonnet's, though its output stays half of Sonnet's.
| Claude Sonnet 5.5 | Grok 4.3 | |
|---|---|---|
| Provider | Anthropic | SpaceXAI |
| Tier | Mid-tier | Mid-tier |
| Input / 1M tokens | $2.00 | $1.25 |
| Output / 1M tokens | $10.00 | $2.50 |
| Cache hit / 1M | $0.10 | $0.20 |
| Cache write / 1M | $2.50 | — |
| Context window | 1M | 1M |
| Notes | Released Sep 28, 2026, replacing Sonnet 5 at the identical $2/$10 per 1M price; Sonnet 5 is now a legacy model | Reasoning workhorse; replaced Grok 4; ≥200K tokens: $2.50/$5, cache $0.40 |
Grok 4.3 is 38% cheaper on input tokens ($1.25 vs $2.00 per 1M). Grok 4.3 is 4.0× cheaper on output tokens ($2.50 vs $10.00 per 1M).
| Workload | Claude Sonnet 5.5 | Grok 4.3 | Difference |
|---|---|---|---|
| Support chatbot 10,000 conversations × (2,000 input + 400 output tokens) | $80.00 | $35.00 | Grok 4.3 saves $45.00 |
| Document Q&A 1,000 questions × (30,000 input + 800 output tokens) | $68.00 | $39.50 | Grok 4.3 saves $28.50 |
| Coding agent session 4M input tokens (90% cache hits) + 150,000 output tokens | $2.66 | $1.59 | Grok 4.3 saves $1.07 |
Standard rates per 1M tokens from data on this site. The coding-agent row assumes 90% of input is read from the prompt cache (full input price where a model has no cache rate). Cache-write fees, batch discounts and long-context surcharges listed in the notes are not included. Updated November 2, 2026.
your evals favour it, most of your input is cached context where its $0.10 cache-hit rate applies, or you need Anthropic's Batch API and enterprise options.
your workload is output-heavy — reasoning traces, long drafts, code generation — and you want the lower per-token cost.
At $2.50 vs $10.00 per 1M output tokens, Grok 4.3 is far cheaper on workloads that generate a lot. On input-heavy document Q&A, the gap is smaller (about 1.7× in the workload table) because the input ratio is narrower.
Anthropic bills Sonnet 5.5 cache hits at 5% of input; Grok 4.3's cache hit is 16% of its input and higher in absolute terms. Sonnet adds a $2.50 write fee per 1M tokens (5-minute cache). On a coding agent that is 90% cached, Grok's output advantage still wins, but by less than the rate card suggests.
xAI bills all tokens in a Grok 4.3 request at double the rate once the prompt reaches 200K tokens, including cached input. Doubled, Grok 4.3's input ($2.50) exceeds Sonnet's $2.00, while its output ($5) stays half of Sonnet's $10.00 — so long, input-heavy prompts can cost more on Grok. Anthropic bills Sonnet 5.5's full 1M window at standard rates; both offer 1M-token windows.
Yes on standard rates: $1.25 vs $2.00 input and $2.50 vs $10.00 output per 1M tokens.
Prompts of 200K tokens or more are billed at double the rate on all tokens. Grok 4.3's cache hit is also higher than Sonnet 5.5's.
Grok 4.3 replaced Grok 4 as xAI's reasoning workhorse.
How we choose: prices come from each vendor's official pricing page and are re-checked every week. We take no affiliate fees, so rankings aren't for sale.