Verified November 2026

Home › Compare › Claude Sonnet 4.6 vs Grok 4.6

Claude Sonnet 4.6 vs Grok 4.6

A straight cost comparison of Claude Sonnet 4.6 (Anthropic) and Grok 4.6 (xAI Grok), per 1,000,000 tokens, verified November 2026.

$3Claude Sonnet 4.6 input / 1M
$2Grok 4.6 input / 1M
48%Blended price gap
Grok 4.6Cheaper blended

Rate-by-rate comparison

USD per 1M tokens · November 2026
MetricClaude Sonnet 4.6Grok 4.6Winner
Input / 1M$3$2Grok 4.6
Cached input / 1M$0.30$0.50Claude Sonnet 4.6
Cache write / 1M$3.75—Grok 4.6
Output / 1M$15$6Grok 4.6
Context window1,000,000500,000Claude Sonnet 4.6
Batch APIYesYesTie
Long-context tierNone> 200,000—

For per-token metrics, the lower number wins. For context window and batch support, the higher or broader option wins. A dash means the provider does not publish that rate: a missing cache-write fee means writes are not billed per token, while a missing cached-input rate means cache hits bill at the full input rate.

Blended cost: 1M input + 250K output

ModelBlended costDifference
Claude Sonnet 4.6$6.75+$3.25
Grok 4.6$3.50Cheaper

Ten workloads, two bills

The same two models, priced across ten workloads that differ only in their input:output ratio. Nothing else changes — no retries, no system prompt, no middleware.

WorkloadClaude Sonnet 4.6Grok 4.6Winner
Classification
100K reqs · 500 in / 10 out
$165$106Grok 4.6
Data extraction
10K reqs · 2K in / 200 out
$90$52Grok 4.6
Support chat
1K reqs · 4K in / 600 out
$21$11.60Grok 4.6
Code review
2K reqs · 8K in / 1.5K out
$93$50Grok 4.6
Summarisation
1K reqs · 20K in / 1K out
$75$46Grok 4.6
Agent loop
5K reqs · 12K in / 2K out
$330$180Grok 4.6
Translation
500 reqs · 15K in / 15K out
$135$60Grok 4.6
Long-document QA
200 reqs · 120K in / 800 out
$74.40$48.96Grok 4.6
Cached RAG
1K reqs · 50K in (45K cached) / 1K out
$43.66$38.57Grok 4.6
Nightly batch
10K reqs · 10K in / 2K out
$300$160Grok 4.6

Claude Sonnet 4.6 wins 0 of 10, Grok 4.6 wins 10 of 10. The winner is set by the token mix, not by the headline rate.

Where the two bills cross over

Grok 4.6 is cheaper on both sides of the mix: $2 vs $3 per million input tokens, and $6 vs $15 per million output. No input:output ratio reverses that, so on price alone the other model never becomes the better choice — it has to win on context, latency or output quality.

Cost at four scales

The same support-chat workload (4K input, 600 output per request) at four monthly volumes.

Requests / monthClaude Sonnet 4.6Grok 4.6Difference
100$2.10$1.16Grok 4.6 cheaper by $0.94
1,000$21$11.60Grok 4.6 cheaper by $9.40
10,000$210$116Grok 4.6 cheaper by $94
100,000$2,100$1,160Grok 4.6 cheaper by $940

What caching does to each bill

Caching only helps the part of the prompt that repeats. This is the support-chat workload (1,000 requests, 4,000 input tokens each) with an increasing share of that prompt served from a warm cache, written once on the first request.

Cached share of the promptClaude Sonnet 4.6Grok 4.6Winner
0%$21$11.60Grok 4.6
50%$15.61$8.60Grok 4.6
80%$12.37$6.80Grok 4.6
95%$10.75$5.91Grok 4.6

Both models publish a cached-read rate, so the discount is comparable: Claude Sonnet 4.6 reads cache at $0.30 per million, Grok 4.6 at $0.50.

Which one should you pick?

Claude Sonnet 4.6 wins on none of the ten workloads above, while Grok 4.6 wins on classification, data extraction, support chat, code review, summarisation, agent loop, translation, long-document qa, cached rag, nightly batch.

Beyond price, the two differ on input rate ($3 vs $2); output rate ($15 vs $6); cached-input rate ($0.30 vs $0.50); cache-write rate ($3.75 vs —); context window (1,000,000 vs 500,000); long-context tier (none vs yes, above 200,000).

Batch mode is available on Claude Sonnet 4.6 and Grok 4.6, which halves the nightly job. Both publish cached-input discounts, so a stable system prompt is worth caching on either.

FAQ

Is Claude Sonnet 4.6 cheaper than Grok 4.6?

On a blended 1M-input-plus-250K-output workload, Grok 4.6 is cheaper by 48% ($3.50 vs $6.75). The gap changes for cache-heavy prompts and for batch jobs, as the workload table shows.

Which handles longer prompts, Claude Sonnet 4.6 or Grok 4.6?

Grok 4.6 has the larger context window (1,000,000 vs 500,000 tokens). Watch the long-context threshold on Grok 4.6: rates rise once input passes it.

Do both models support prompt caching?

Claude Sonnet 4.6 bills cached reads at $0.30 per 1M; Grok 4.6 at $0.50.

Related comparisons

Run your own token volumes