Welcome to the February edition of our monthly LLM rankings. Each month, we pull real-world usage data from OpenRouter — a platform routing millions of API calls across every major LLM — and rank the models developers are actually paying for.
Not benchmarks. Not press releases. Real tokens, real money, real usage.
The Top 10 — February 2026
| Rank | Model | Provider | Move | Input / 1K tokens | Output / 1K tokens |
|---|---|---|---|---|---|
| 1 | Claude 3.5 Sonnet | Anthropic | ● | $0.003 | $0.015 |
| 2 | Claude Sonnet 4.6 | Anthropic | 🆕 | $0.003 | $0.015 |
| 3 | Gemini 2.0 Flash | ↓ | $0.0001 | $0.0004 | |
| 4 | GPT-5.2 | OpenAI | ↑ | $0.00175 | $0.014 |
| 5 | DeepSeek V3 | DeepSeek | ↓ | $0.0003 | $0.0009 |
| 6 | Gemini 3 Flash | 🆕 | $0.0001 | $0.0004 | |
| 7 | GPT-4o | OpenAI | ↓ | $0.005 | $0.015 |
| 8 | Claude Opus 4.6 | Anthropic | 🆕 | $0.015 | $0.075 |
| 9 | DeepSeek R1 | DeepSeek | ● | $0.0006 | $0.002 |
| 10 | Qwen3 Max Thinking | Alibaba | 🆕 | $0.0002 | $0.001 |
New This Month
Claude Sonnet 4.6 (Jan 28) — Anthropic's mid-cycle upgrade slots in right behind 3.5 Sonnet at the same price point. Better instruction following, stronger agentic tool use, and noticeably improved code editing. Developers are migrating fast — the question isn't if it overtakes 3.5, but when.
Claude Opus 4.6 (Jan 28) — The heavy hitter. Opus pricing keeps it niche, but for complex reasoning chains, legal analysis, and multi-step research tasks, nothing else comes close. Enterprise adoption is strong.
Gemini 3 Flash (Jan 15, GA) — Google's next-gen speed model graduates from preview to general availability. Same absurd throughput as the preview, now with production SLAs. Already eating Gemini 2.0 Flash's volume from the bottom.
Qwen3 Max Thinking (Jan 22) — Alibaba's reasoning-first model enters the top 10 on price alone. At $0.0002/1K input tokens, it's the cheapest thinking model on the market. Quality lags behind Claude and GPT for English tasks, but it's dominant for Chinese-language workloads.
The Story
February's headline: Anthropic owns three of the top ten spots. Claude 3.5 Sonnet holds the crown for the second consecutive month, but Claude Sonnet 4.6 is climbing so fast that the two might swap positions by March. Developers with existing 3.5 Sonnet integrations are cautious — the models share identical pricing, but 4.6's behavioral differences require testing.
GPT-5.2 is the month's biggest mover, jumping three spots to #4 as developers finish evaluating and start committing production traffic. OpenAI's pricing undercut worked.
Google's Flash lineup now holds two top-10 spots. The combined Flash volume (2.0 + 3) makes Google the clear leader in high-throughput, low-cost inference. If you're processing millions of documents, you're probably running Flash.
GPT-4o drops to #7 — its first time outside the top 5. The writing is on the wall for last-gen flagships.
Market Share
| Provider | Share | Trend |
|---|---|---|
| Anthropic | ~34% | Growing |
| ~24% | Stable | |
| OpenAI | ~19% | Declining |
| DeepSeek | ~11% | Stable |
| Alibaba | ~4% | New entry |
| Meta | ~4% | Declining |
| Mistral | ~3% | Stable |
| Others | ~1% | — |
Looking Ahead
Watch for Gemini 3.1 Pro rumors — Google I/O season is approaching and leaks suggest a major model upgrade. Claude Sonnet 4.6 will likely overtake 3.5 Sonnet next month as migration completes. And DeepSeek V4 is reportedly in testing.
Monthly LLM rankings by Demand Signals, sourced from OpenRouter usage data. Subscribe to our blog for monthly updates.
Get a Free AI Demand Gen Audit
We'll analyze your current visibility across Google, AI assistants, and local directories — and show you exactly where the gaps are.