Compare AI Models Side by Side
Pick 2–4 models to compare Agent Arena performance, context windows, and the cheapest API token prices across every channel.
Current Available API Leaders by Vendor
One currently purchasable general-purpose model per vendor. A model explicitly designated as the flagship is preferred; otherwise selection uses Agent Arena, then release date when no comparable score exists. Add a model to the comparison or inspect every channel offering it.
| Vendor | Available model | Released | Agent Score | Context Window | Cheapest Input / 1M | Cheapest Output / 1M | Cheapest channel | Channel prices |
|---|---|---|---|---|---|---|---|---|
| OpenAI | Sep 5, 2026 | 12.55% | 1.1M | $0.88 | $5.25 | XiuRouter | Compare 4 channels | |
| Anthropic | Sep 2, 2026 | 14.51% | 1M | $2.30 | $11.50 | XiuRouter | Compare 4 channels | |
| Gemini(Google) | Sep 2, 2026 | 4.01% | 1M | $0.38 | $1.88 | OpenRouter | Compare 4 channels | |
| Grok / X.AI | Jul 8, 2026 | 3.93% | 500K | $0.20 | $0.60 | XiuRouter | Compare 3 channels | |
| DeepSeek | Aug 12, 2026 | 4.43% | 1M | $0.39 | $0.78 | XiuRouter | Compare 6 channels | |
| 智谱 AI (ChatGLM) | Jun 17, 2026 | 4.64% | 1M | $0.70 | $2.20 | OpenRouter | Compare 5 channels | |
| 月之暗面 (Moonshot/Kimi) | Jul 16, 2026 | 6.59% | 128K | $2.34 | $11.70 | OpenRouter | Compare 5 channels | |
| Qwen(阿里) | Aug 3, 2026 | 3.9% | 1M | ¥12.00 | ¥36.00 | Qwen(阿里) | Compare 2 channels | |
| Minimax | Jun 1, 2026 | -5.26% | 1M | $0.30 | $1.20 | OpenRouter | Compare 5 channels |
Choose 2–4 models. Selecting a 5th replaces the first one.
Gemini(Google)
Anthropic
| Compare AI Models Side by Side | Gemini 3.8 Flash | Claude Fable 5.1 |
|---|---|---|
| Performance | ||
| Agent Arena score | 4.01% | 14.51%Best |
| GPQA DiamondArtificial Analysis current suite · Diamond · ACCURACY | 95.3%Best | 93.7% |
| Humanity's Last ExamArtificial Analysis current suite · Text only · ACCURACY | 47.8% | 59.1%Best |
| SciCodeArtificial Analysis current suite · Main · PASS_RATE | 56.6% | 63.1%Best |
| MMMU-ProArtificial Analysis current suite · Main · ACCURACY | 85.6% | — |
| Context & Limits | ||
| Context window | 1MLargest | 1M |
| API Pricing (per 1M tokens) | ||
| Cheapest input | $0.38Cheapest | $2.30 |
| Cheapest output | $1.88Cheapest | $11.50 |
| vs Official | -50% | -77% |
| Links | ||
| View all channel prices | Details | Details |
| Compare subscription plans | Compare Plans | Compare Plans |
At a glance
Winners within your current selection.
Frequently asked questions
Which AI model is the best value for API usage?
It depends on your task. For high-volume, latency-tolerant workloads, open models like DeepSeek and Qwen on aggregator channels are often the cheapest per 1M tokens; frontier models like Claude Opus and the GPT series cost more but lead on Agent Arena performance. Use this page to compare the cheapest available channel price of each model side by side — prices are normalised to USD for comparison but displayed in each channel’s own currency.
How are API token prices compared across channels and currencies?
Every channel price is converted to USD using cached exchange rates before taking the minimum, so a ¥3/1M CNY channel is correctly ranked against a $0.78/1M USD channel. The cheapest input price determines the recommended channel; output and cached-input prices are shown alongside. The "vs Official" row shows how much cheaper (or dearer) that channel is versus the model producer’s own API.
How many models can I compare at once?
You can compare between two and four models side by side. Pick them from the search box, or share the URL — the ?models= slug list preselects the same comparison for whoever opens it.