LLM API Provider Pricing Comparison
Published 2026-06-25 · Data from 2137 chat models across 5 providers
Picking the right LLM provider for your application depends on more than just raw pricing — but pricing is where most decisions start. This guide compares OpenAI, Anthropic, Google, Mistral, and DeepSeek across multiple pricing tiers to help you find the best fit for your workload and budget.
- Which provider has the cheapest chat model?
- Best mid-range options: quality vs price
- Cheapest reasoning models
- Provider pricing averages and trends
- Summary: which provider should you choose?
| Provider | Model | Input / 1M | Output / 1M | 100k messages/mo (USD) | Context | Features |
|---|---|---|---|---|---|---|
| Mistral | ministral-3-3b-2512 | $0.1000 / 1M tokens | $0.1000 / 1M tokens | $8.0000 | 131.1K | Vision, Tools |
| gemini-2.0-flash-lite-001 | $0.0750 / 1M tokens | $0.3000 / 1M tokens | $12.7500 | N/A | Vision, Tools, Cache | |
| DeepSeek | deepseek-coder | $0.1400 / 1M tokens | $0.2800 / 1M tokens | $15.4000 | 4.1K | Tools, Cache |
| OpenAI | gpt-5-nano-2025-08-07 | $0.0500 / 1M tokens | $0.4000 / 1M tokens | $14.5000 | 128.0K | Vision, Tools, Reasoning, Cache |
| Anthropic | Claude Haiku 5.5 | $0.1000 / 1M tokens | $0.5000 / 1M tokens | $20.0000 | 1.0M | Vision, Tools, Reasoning, Cache |
The 100k-message monthly example assumes 500 input and 300 output tokens per message (50M input and 30M output tokens total). Each figure uses the listed model's USD rates: 50 × input price per 1M + 30 × output price per 1M. It excludes caching or batch discounts, routing, tool calls, taxes, and other charges. Availability and pricing can change; this example is not a quality ranking, a billing guarantee, or a claim that the displayed model is the cheapest for this workload.
Mistral has the lowest combined rate for 1M input plus 1M output tokens among these five providers: $0.2000 with ministral-3-3b-2512. For output-heavy workloads, Mistral has the lowest listed output rate at $0.1000 / 1M output tokens with ministral-3-3b-2512. Actual request costs depend on the input/output token mix.
For a quick price check across two models, use the compare tool. To adjust the monthly example above, open the calculator with 500 input tokens, 300 output tokens, and 100,000 requests prefilled.
Best mid-range options: quality vs price
The absolute cheapest models are usually small or older. When you need a balance of capability, context length, and cost, per-provider mid-range options offer better performance per dollar:
| Provider | Model | Input / 1M | Output / 1M | Context |
|---|---|---|---|---|
| OpenAI | ft:gpt-4o-mini-2024-07-18 | $0.3000 / 1M tokens | $1.2000 / 1M tokens | 16.4K |
| Anthropic | claude-haiku-4-5-20251001 | $1.0000 / 1M tokens | $5.0000 / 1M tokens | 64.0K |
| gemini-robotics-er-1.5-preview | $0.3000 / 1M tokens | $2.5000 / 1M tokens | 65.5K | |
| Mistral | codestral-2508 | $0.3000 / 1M tokens | $0.9000 / 1M tokens | 256.0K |
| DeepSeek | deepseek-v4-pro | $0.4350 / 1M tokens | $0.8700 / 1M tokens | 8.2K |
Cheapest reasoning models
Reasoning (chain-of-thought) models are increasingly popular for complex logic and coding. Here are the cheapest per-provider options with reasoning capabilities:
| Provider | Model | Input / 1M | Output / 1M | Context |
|---|---|---|---|---|
| OpenAI | gpt-5-nano-2025-08-07 | $0.0500 / 1M tokens | $0.4000 / 1M tokens | 128.0K |
| Anthropic | Claude Haiku 5.5 | $0.1000 / 1M tokens | $0.5000 / 1M tokens | 1.0M |
| gemini-flash-lite-latest | $0.1000 / 1M tokens | $0.4000 / 1M tokens | 65.5K | |
| Mistral | mistral-small-2603 | $0.1500 / 1M tokens | $0.6000 / 1M tokens | 262.1K |
| DeepSeek | deepseek-v3.2 | $0.2800 / 1M tokens | $0.4000 / 1M tokens | 163.8K |
Provider pricing averages and trends
These averages treat every listed, USD-priced chat model equally, rather than weighting by actual usage. They describe this catalog sample, not the price of a typical request:
| Provider | Chat models | Avg input / 1M | Avg output / 1M | Cheapest combined |
|---|---|---|---|---|
| DeepSeek | 10 | $0.3230 / 1M tokens | $0.7390 / 1M tokens | $0.4200 / 1M tokens |
| 46 | $0.7380 / 1M tokens | $4.5815 / 1M tokens | $0.3750 / 1M tokens | |
| Mistral | 56 | $0.8486 / 1M tokens | $2.6332 / 1M tokens | $0.2000 / 1M tokens |
| OpenAI | 96 | $3.8224 / 1M tokens | $13.7146 / 1M tokens | $0.4500 / 1M tokens |
| Anthropic | 32 | $6.2297 / 1M tokens | $31.1484 / 1M tokens | $0.6000 / 1M tokens |
DeepSeek has the lowest unweighted average input price at $0.3230 / 1M tokens / 1M, followed by Google at $0.7380 / 1M tokens / 1M. Compare individual models and workload token ratios before selecting a provider.
Summary: which provider should you choose?
| Use case | Provider | Example model | Why |
|---|---|---|---|
| Lowest equal-weight token rate | Mistral | ministral-3-3b-2512 | Lowest sum of 1M input + 1M output token rates |
| Lowest output-token rate | Mistral | ministral-3-3b-2512 | Lowest listed price per 1M output tokens |
| Lowest reasoning-model token rate | OpenAI | gpt-5-nano-2025-08-07 | Lowest combined rates among models tagged with reasoning |
| Quality-sensitive workloads | Anthropic / OpenAI | Compare GPT-5.2 vs Claude Sonnet 4.6 | Compare like-for-like benchmarks and constraints before choosing |
Every workload is different — use the model directory with benchmark, pricing, and feature filters to find the exact fit for your use case, or load a calculator preset to estimate your monthly cost.
Pricing data sourced from our catalog. Check data sources for provenance and freshness.