// head_to_head

Claude Sonnet 5 vs Gemini 3.1 Pro Preview

Published LiveBench scores across all seven categories, live list pricing, context windows, and the measured cost of a point of capability — for both models, side by side.

OpenRouter + LiveBenchAll comparisonsFull leaderboard

Effectively the same quality — Claude Sonnet 5 is the cheaper way to get it.

The two are within 0.9 points overall, which is inside the range that effort settings alone move a LiveBench score, so treat them as quality-equivalent. Claude Sonnet 5 lists 13% cheaper per blended million tokens. When quality ties, cost is the whole decision. The two cost measures disagree here, which is worth knowing: Claude Sonnet 5 has the lower sticker price, but Gemini 3.1 Pro Preview earns each point of capability for less — $0.1567 against $0.2691 — because per-token rates do not predict how many tokens a model actually spends on a task.

anthropic

Claude Sonnet 5

Blended / 1M
$4.00
Context
1M
Released
Jun 30, 2026
Overall score
76.0
reasoningtool callingimage inputfile inputprompt caching

google

Gemini 3.1 Pro Preview

Blended / 1M
$4.50
Context
1.0M
Released
Feb 19, 2026
Overall score
77.0
reasoningtool callingaudio inputfile inputimage inputvideo inputprompt caching

Specs and pricing

MetricClaude Sonnet 5Gemini 3.1 Pro Preview
LiveBench overall

Mean of the seven LiveBench category scores, 0–100. Higher is better.

76.077.0
Cost per point

Measured benchmark spend divided by overall score — dollars per point of capability.

$0.2691$0.1567win
Blended price / 1M

3:1 input:output mix, the usual shape of production traffic.

$4.00win$4.50
Input price / 1M$2.00$2.00
Output price / 1M$10.00win$12.00
Cached input / 1M

Price of an input token served from the prompt cache, where the provider publishes one.

$0.200$0.200
Context window1M1.0M
Max output tokens128Kwin66K

Benchmarks by category

An overall score averages away the thing you probably care about. These are the seven LiveBench categories scored separately — Claude Sonnet 5 on top, Gemini 3.1 Pro Preview below, both out of 100.

Agentic coding
59.4
44.1
Coding
80.7
76.5
Reasoning
88.7
84.0
Mathematics
92.9
91.0
Data analysis
71.7
78.5
Language
75.0
85.4
Instruction following
63.9
79.1

What each one costs to run

Per-token prices are hard to feel. These are monthly list costs for both models across five workload shapes, using each provider's published cached-input rate where there is one.

WorkloadClaude Sonnet 5Gemini 3.1 Pro Preview
Support chatbot

1.2K in / 400 out × 200K requests

$1150.40/mo$1310.40/mo
RAG assistant

8K in / 600 out × 100K requests

$1480.00/mo$1600.00/mo
Coding agent

40K in / 4K out × 20K requests

$1392.00/mo$1552.00/mo
Document extraction

20K in / 1.5K out × 50K requests

$2660.00/mo$2810.00/mo
Bulk classification

500 in / 20 out × 5M requests

$5100.00/mo$5300.00/mo
Run these two through the cost calculator

Which should you pick?

You are running this at volume

Gemini 3.1 Pro Preview

Lowest measured cost per point of capability at $0.1567 per point — the gap compounds with every request.

The workload is coding or agentic work

Claude Sonnet 5

Leads on agentic coding — 59.4 against 44.1.

Claude Sonnet 5 vs Gemini 3.1 Pro Preview FAQ

Which is better, Claude Sonnet 5 or Gemini 3.1 Pro Preview?

Effectively the same quality — Claude Sonnet 5 is the cheaper way to get it. The two are within 0.9 points overall, which is inside the range that effort settings alone move a LiveBench score, so treat them as quality-equivalent. Claude Sonnet 5 lists 13% cheaper per blended million tokens. When quality ties, cost is the whole decision. The two cost measures disagree here, which is worth knowing: Claude Sonnet 5 has the lower sticker price, but Gemini 3.1 Pro Preview earns each point of capability for less — $0.1567 against $0.2691 — because per-token rates do not predict how many tokens a model actually spends on a task.

Is Claude Sonnet 5 cheaper than Gemini 3.1 Pro Preview?

Claude Sonnet 5 is cheaper. On a 3:1 input:output blend, Claude Sonnet 5 lists at $4.00 per million tokens and Gemini 3.1 Pro Preview at $4.50 — Claude Sonnet 5 is 13% cheaper. Input and output are priced separately — Claude Sonnet 5 charges $2.00 in and $10.00 out, Gemini 3.1 Pro Preview charges $2.00 and $12.00 — so the model that looks cheaper flips depending on how output-heavy your workload is.

Claude Sonnet 5 vs Gemini 3.1 Pro Preview: which scores higher on benchmarks?

Claude Sonnet 5 scores 76.0 and Gemini 3.1 Pro Preview scores 77.0 overall on LiveBench, the mean of its seven categories. That gap is inside the range that effort settings alone move a score, so treat them as equivalent on published quality. Category scores differ from the overall figure — a model can lead on reasoning and trail on coding, which the per-category table above breaks out.

Which gives better value for money, Claude Sonnet 5 or Gemini 3.1 Pro Preview?

Gemini 3.1 Pro Preview. Cost per point divides the measured dollars LiveBench spent running the benchmark by the score it earned, so it captures something token pricing misses: a reasoning model can emit many times more tokens than its per-token rate suggests. Claude Sonnet 5 works out at $0.2691 per point and Gemini 3.1 Pro Preview at $0.1567.

Does Claude Sonnet 5 or Gemini 3.1 Pro Preview have a bigger context window?

They are effectively the same — 1M for Claude Sonnet 5 and 1.0M for Gemini 3.1 Pro Preview.

Do Claude Sonnet 5 and Gemini 3.1 Pro Preview support prompt caching?

Both publish a cached-input rate: $0.200 per million for Claude Sonnet 5 and $0.200 for Gemini 3.1 Pro Preview, against full input rates of $2.00 and $2.00. On a workload with a long stable prefix — a system prompt, a tool schema, a retrieved corpus — that changes the economics more than the headline price does.

Related comparisons

How these numbers are produced

  • Price — provider list price from OpenRouter, refreshed every 15 minutes. “Blended” is a 3:1 input:output mix.
  • ScoresLiveBench release 2026-06-25, using their own category map. Each model shows its strongest published run. A blank means “not evaluated”, never “bad”.
  • Cost per point — the measured dollars LiveBench spent on the run, divided by the score it earned.
  • “Win” — awarded only past a threshold: one full point on a benchmark score, 10% on a price, 25% on a context window. Anything tighter reports as a tie, because effort settings alone move a LiveBench score by more than that.

Published benchmarks rank models on someone else's tasks. Before committing, see LLM & agent evaluation for building an eval on your own.