Sub-$2 flagships with million-token windows: Google tiered vs Anthropic flat.
gemini-3.1-pro · google
claude-sonnet-5 · anthropic
Benchmark
| Workload Scenario | Gemini 3.1 Pro | Claude Sonnet 5 | Price Delta |
|---|---|---|---|
| 1M input tokens (raw text) | $2.00 | $2.00 | — |
| 1M output tokens (generation) | $12.00 | $10.00 | +$2.00 |
| 1M tokens · 70% input / 30% output mix | $5.00 | $4.40 | +$0.60 |
| Standard chat turn (4K in / 800 out, 50% cached) | $0.014 | $0.0124 | +$0.0016 |
| Monthly scale (10K requests / day) | $4,200.00 | $3,720.00 | +$480.00 |
Negative difference = Gemini 3.1 Pro is cheaper. Positive = Claude Sonnet 5 is cheaper.
Scaling Curve
Total cost of a token volume at 70/30 input/output split (uncached). Log-log scale.
Verdict
Claude Sonnet 5 is cheaper on both input and output rates, so it costs less at every input/output mix. Price alone still isn't the whole decision: capability, latency and context limits (Gemini 3.1 Pro: 1M, Claude Sonnet 5: 1M) may justify the premium for your task.
FAQ
On input, Claude Sonnet 5 is cheaper ($2.00/M vs $2.00/M). On output, Claude Sonnet 5 is cheaper ($10.00/M vs $12.00/M). The same model is cheaper on both sides, so it wins at every mix.
A chat-style request (4,000 input + 800 output tokens, 50% cached) costs $0.014 on Gemini 3.1 Pro and $0.0124 on Claude Sonnet 5 — Claude Sonnet 5 is 1.1× more expensive for that workload.
Gemini 3.1 Pro supports 1,048,576 tokens (65.5K max output); Claude Sonnet 5 supports 1,000,000 (128K max output). Gemini 3.1 Pro fits 1.0× more context, which matters for long documents and agents.