Opus 5 vs GPT-5.6 Sol vs Gemini 3.1 Pro
The three Western flagships. All three claim a million-token context — and all three charge you differently for actually using it.
Updated August 28, 2026 · Independent comparison — Claude Opus 5, GPT-5.6 Sol, Gemini 3.1 Pro are separate products.
Claude Opus 5
Anthropic's agentic coding default
Claude Opus 5 launched on 2026-07-24 at $5 in and $25 out per million tokens, with cache reads at $0.50 and a 50% batch discount. It offers a 1M-token context with 128K output (300K in the Batch API beta), and Anthropic positions it as the default for complex agentic coding. It is available as claude-opus-5 on the Claude API, Bedrock, Vertex and Microsoft Foundry, inside Claude Code on Pro and Max, and through gateways such as OpenCode Zen at the same $5/$25.
GPT-5.6 Sol
OpenAI's deep-reasoning tier
GPT-5.6 Sol is OpenAI's deep-reasoning tier, currently on promotional pricing of $4 in and $20 out per million tokens through at least 2026-11-21, with cached input at $0.40. Context runs to about 1.1M, but requests above 272K input tokens are surcharged at 2x input and 1.5x output for the whole request. The alias gpt-5.6 resolves to Sol. It requires ChatGPT Plus or above on the consumer side; Fast mode costs 2x.
Gemini 3.1 Pro
Google's paid flagship, tiered by context
Gemini 3.1 Pro prices by context rather than flat: $2 in and $12 out per million tokens at or below 200K, rising to $4 and $18 above it, with caching at $0.20/$0.40 and cache storage billed at $4.50 per million tokens per hour. Context is 1M with 65,536 output tokens. There is no free tier on this SKU — it is paid Gemini API only — though it also appears in Antigravity, AI Studio and Vertex, and Batch or Flex processing runs about 50% cheaper.
Bottom line
All three are million-token flagships, and the sticker price is misleading because each one surcharges long context differently. Claude Opus 5 is $5 in and $25 out per million tokens flat, with 128K output and a 50% batch discount — the simplest pricing of the three. GPT-5.6 Sol is currently $4/$20 on a promotion running to at least 2026-11-21, but any request above 272K input tokens is billed at 2x input and 1.5x output for the entire request, which makes genuinely long context expensive. Gemini 3.1 Pro tiers explicitly: $2/$12 at or below 200K and $4/$18 above, plus cache storage at $4.50 per million tokens per hour, and it has the smallest output ceiling at 65,536 tokens.
Pricing
| Feature | Claude Opus 5 | GPT-5.6 Sol | Gemini 3.1 Pro |
|---|---|---|---|
| Price per Mtok (in / out) | $5 in / $25 out per Mtok | $4 in / $20 out per Mtok (promo to at least 2026-11-21) | $2/$12 per Mtok at or below 200K; $4/$18 above |
| Cache pricing | $0.50 read; $6.25-$10 write; 50% batch discount | $0.40 cached input; $5 cache write | $0.20-$0.40; storage $4.50 per Mtok per hour |
Capacity
| Feature | Claude Opus 5 | GPT-5.6 Sol | Gemini 3.1 Pro |
|---|---|---|---|
| Context window | 1M | About 1.1M | 1M |
| Max output tokens | 128K (300K in Batch API beta) | Not published | 65,536 |
Availability
| Feature | Claude Opus 5 | GPT-5.6 Sol | Gemini 3.1 Pro |
|---|---|---|---|
| Open weights | ✗ | ✗ | ✗ |
| Where you can run it | Claude API, Bedrock, Vertex, Microsoft Foundry, Claude Code, OpenCode Zen, DevPass | OpenAI API, ChatGPT Plus and above, Codex, Cursor, OpenCode Zen | Paid Gemini API, Antigravity, AI Studio, Vertex |
| Released | 2026-07-24 | Current GPT-5.6 flagship | gemini-3.1-pro-preview |
Verdict at a glance
| Feature | Claude Opus 5 | GPT-5.6 Sol | Gemini 3.1 Pro |
|---|---|---|---|
| Best for | Complex agentic coding where quality outweighs cost | Deep reasoning where you can avoid the 272K surcharge | Long-context work where you stay under 200K per request |
The Verdict
Choose Claude Opus 5 if...
- ✓You want the simplest pricing — one rate regardless of context length
- ✓You want Anthropic's agentic coding quality and a 128K output ceiling
- ✓Batch processing at 50% off fits part of your workload
- ✓You want the broadest enterprise availability: Bedrock, Vertex and Microsoft Foundry
Choose GPT-5.6 Sol if...
- ✓You can keep requests under 272K input and take the promotional $4/$20
- ✓You want the largest raw context here at about 1.1M
- ✓You are already inside the OpenAI ecosystem via ChatGPT or Codex
- ✓You want deep reasoning as the model's primary strength
Choose Gemini 3.1 Pro if...
- ✓Your requests stay under 200K, where $2/$12 is the cheapest of the three
- ✓You want explicit, predictable context tiering rather than a whole-request surcharge
- ✓Batch or Flex processing at about 50% off suits your workload
- ✓You can live with a 65,536-token output ceiling
Frequently Asked Questions
What is the difference between Claude Opus 5, GPT-5.6 Sol and Gemini 3.1 Pro?
Claude Opus 5 launched on 2026-07-24 at $5 in and $25 out per million tokens, with cache reads at $0.50 and a 50% batch discount. GPT-5.6 Sol is OpenAI's deep-reasoning tier, currently on promotional pricing of $4 in and $20 out per million tokens through at least 2026-11-21, with cached input at $0.40. Gemini 3.1 Pro prices by context rather than flat: $2 in and $12 out per million tokens at or below 200K, rising to $4 and $18 above it, with caching at $0.20/$0.40 and cache storage billed at $4.50 per million tokens per hour.
How much do Claude Opus 5, GPT-5.6 Sol and Gemini 3.1 Pro cost?
Claude Opus 5: $5 in / $25 out per Mtok. GPT-5.6 Sol: $4 in / $20 out per Mtok (promo to at least 2026-11-21). Gemini 3.1 Pro: $2/$12 per Mtok at or below 200K; $4/$18 above. See the pricing table above for full plan details.
Which of Claude Opus 5, GPT-5.6 Sol and Gemini 3.1 Pro should I pick?
Choose Claude Opus 5 if you want the simplest pricing — one rate regardless of context length. Choose GPT-5.6 Sol if you can keep requests under 272K input and take the promotional $4/$20. Choose Gemini 3.1 Pro if your requests stay under 200K, where $2/$12 is the cheapest of the three.
1DevToolSwitch models without switching tools
Picking a model is a decision you will make again in three months, so the thing worth optimising is how cheaply you can change your mind. 1DevTool runs Claude Code, Codex CLI, Cline, OpenCode and Gemini CLI side by side in persistent terminals — point each at whichever model is winning this quarter, keep your own keys, and get an HTTP client, a 13-engine database client and an embedded browser in the same window. One-time $29, no subscription.