High-volume, cost-sensitive work
Either one
Both cost $4.00 per million tokens on a typical mix of three input tokens for every output token.
Choose Claude Sonnet 5 for reasoning over long documents. Choose Gemini 4 Argon for reused prompts and documents (RAG, agents), the hardest reasoning tasks and the latest release.
Each answer below comes straight from the published prices, specifications and independent benchmark results.
Either one
Both cost $4.00 per million tokens on a typical mix of three input tokens for every output token.
Choose Gemini 4 Argon
Gemini 4 Argon charges $0.10 per million cached input tokens, against $0.20 for Claude Sonnet 5.
Choose Gemini 4 Argon
Gemini 4 Argon scores 52.6 on the Artificial Analysis Intelligence Index, against 38.2 for Claude Sonnet 5.
Choose Claude Sonnet 5
Claude Sonnet 5 scores 82% on Long-context reasoning (AA-LCR), against 79.7% for Gemini 4 Argon.
Choose Gemini 4 Argon
Gemini 4 Argon is the newer release (October 2026, against June 2026 for Claude Sonnet 5).
| Volume | Claude Sonnet 5 | Gemini 4 Argon | Difference |
|---|---|---|---|
| Prototype: 1M input, 250K output tokens | $4.50 | $4.50 | = |
| Production: 20M input, 5M output tokens | $90.00 | $90.00 | = |
| Scale: 500M input, 100M output tokens | $2,000 | $2,000 | = |
Standard rates, without caching or batch discounts. Volumes are per month.
| Claude Sonnet 5 | Gemini 4 Argon | |
|---|---|---|
| Provider | Anthropic | |
| Category | Language models | Language models |
| Status | Legacy | Generally available |
| Headline price | $4.00 / 1M tokens | $4.00 / 1M tokens |
| Input | $2.00 / 1M tokens | $2.00 / 1M tokens |
| Output | $10.00 / 1M tokens | $10.00 / 1M tokens |
| Cached input | $0.20 / 1M tokens | $0.10 / 1M tokens |
| Context window | 1M | 1.05M |
| Max output | 128k | 66k |
| Released | June 2026 | October 2026 |
| Knowledge cutoff | No data | No data |
| Licence | Proprietary | Proprietary |
| Capabilities | Vision, Reasoning, Tool calling, Structured output, Prompt caching, Batch API | Vision, Reasoning, Tool calling, Structured output, Prompt caching, Batch API |
| API model id | claude-sonnet-5 | gemini-4-argon |
Leader in bold.
| Benchmark | Claude Sonnet 5 | Gemini 4 Argon |
|---|---|---|
| Intelligence Index | 38.2 | 52.6 |
| SciCode | 54.3% | 61.8% |
| Humanity's Last Exam | 41.3% | 57.1% |
| Long-context reasoning (AA-LCR) | 82% | 79.7% |
| GPQA Diamond | 91.1% | No data |
| MMMU-Pro (visual reasoning) | 77.3% | No data |
Independent evaluations by Artificial Analysis, run the same way for every model, at the reasoning setting shown. Percentages unless marked as an index.
Choose Claude Sonnet 5 for reasoning over long documents. Choose Gemini 4 Argon for reused prompts and documents (RAG, agents), the hardest reasoning tasks and the latest release.
Both cost the same at the headline price ($4.00 / 1M tokens).
Gemini 4 Argon reads more per request: 1.05M tokens against 1M.
Claude Sonnet 5 is made by Anthropic and Gemini 4 Argon by Google. Both cost the same at the headline price ($4.00 / 1M tokens). Gemini 4 Argon reads more per request: 1.05M tokens against 1M. Gemini 4 Argon is the newer model (October 2026, against June 2026).
At 20 million input and 5 million output tokens a month, Claude Sonnet 5 costs $90.00 and Gemini 4 Argon costs $90.00, at standard rates.
Gemini 4 Argon was released on October 2026 and Claude Sonnet 5 on June 2026.