Claude Sonnet 5 vs Gemini 3 Pro

Claude Sonnet 5 and Gemini 3 Pro sit close in price but differ on availability. Sonnet 5 costs $2 input and $10 output per million tokens through August 2026, rising to $3 and $15 on September 1, 2026; Gemini 3 Pro costs $2 input and $12 output, with $0.20 cached input. Sonnet 5 is generally available; Gemini 3 Pro is still in preview.

Anthropic released Sonnet 5 on June 30, 2026 for frontier performance across coding, agents, and professional work at scale. Google positions Gemini 3 Pro as a capable, cost-efficient Pro-tier model.

This page compares price, availability, and compliance so you can pick the right model for your workload.

Claude Sonnet 5 vs. Gemini 3 Pro: Side-by-Side

Price: nearly tied, Sonnet 5's rises in September

The two models are close on price today. Claude Sonnet 5 costs $2 input and $10 output per million tokens through August 2026; Gemini 3 Pro costs $2 input and $12 output, with $0.20 cached input.

That changes on September 1, 2026, when Sonnet 5's intro pricing ends and it moves to $3 input and $15 output per million tokens (Anthropic) — a bit more expensive than Gemini 3 Pro on both sides from that date forward.

Factor the September price change into any long-term budget comparison, not just today's rates.

• Claude Sonnet 5: $2 / $10 through Aug 2026, then $3 / $15 from Sep 1, 2026.
• Gemini 3 Pro: $2 / $12 per million tokens, $0.20 cached input.
• Nearly tied today; Gemini 3 Pro becomes cheaper after Sonnet 5's September price rise.

→ Compare per-token pricing for every AI model

Availability: Sonnet 5 is GA, Gemini 3 Pro is preview

Claude Sonnet 5 is generally available now, released June 30, 2026. Gemini 3 Pro remains in preview status as of this writing.

For production workloads where stability and support commitments matter, that access gap favors Sonnet 5 today. Confirm Gemini 3 Pro's current production guidance with Google before routing committed traffic to it.

• Sonnet 5: generally available since June 30, 2026.
• Gemini 3 Pro: preview status, confirm production-readiness with Google.

Claude Sonnet 5 vs Gemini 3 Pro benchmarks: what's actually published

Neither Claude Sonnet 5's nor Gemini 3 Pro's maker has published a same-day, apples-to-apples benchmark suite pitting these two exact versions against each other, so treat any comparison here as directional rather than a verified head-to-head.

Vendors have not published a directly comparable, head-to-head SWE-bench Verified score for these exact current-generation versions as of this writing. The most recent published reference points are Claude Opus 4.8 at about 88.6%, GPT-5 at about 74%, Grok 4 at about 75%, and Gemini 2.5 Pro at about 71% (Google) — treat these as directional, not a same-generation apples-to-apples test.

Until both vendors publish a matched benchmark run, the more reliable signal is a short pilot on your own tasks — coding, drafting, or analysis representative of your real workload.

• Claude Opus 4.8: ~88.6% on SWE-bench Verified.
• GPT-5: ~74% on SWE-bench Verified.
• Grok 4: ~75% on SWE-bench Verified.
• Gemini 2.5 Pro: ~71% on SWE-bench Verified.

Compliance: Sonnet 5 has documented coverage today

Claude Sonnet 5 inherits Anthropic's SOC 2, ISO 27001, and HIPAA BAA coverage on the API and Enterprise plans.

Gemini 3 Pro's compliance documentation for its preview release was not detailed in materials reviewed for this comparison. Confirm current certification and data-handling terms directly with Google before sending regulated data.

• Sonnet 5: SOC 2, ISO 27001, HIPAA BAA.
• Gemini 3 Pro: confirm preview-stage compliance terms directly with Google.

How to choose for your business

Choose Claude Sonnet 5 if you need a generally available model with documented compliance today, especially for multi-agent, production-scale work. Choose Gemini 3 Pro if you're comfortable piloting a preview-stage model and want to lock in its current per-token price before Sonnet 5's September increase.

Run a short pilot on your own workload before committing production traffic to either model.

• Production-ready today with documented compliance: Claude Sonnet 5.
• Piloting a preview model, cost-sensitive after September: Gemini 3 Pro.

The Verdict

Claude Sonnet 5 is the safer default today: it's generally available, documented for compliance, and built for multi-agent, enterprise-scale work. Gemini 3 Pro is competitively priced but remains in preview as of this writing.

Factor in Sonnet 5's September 1, 2026 price increase — after that date, Gemini 3 Pro becomes the cheaper of the two on a per-token basis, if it has reached general availability by then.