AI Model Pricing: April 2026 Update
April 04, 2026
The AI model pricing landscape continues to evolve rapidly. This April 2026 update covers pricing movements across 891 language models from 64 providers that we track daily.
Whether you are building AI-powered applications, evaluating providers for enterprise deployment, or simply keeping an eye on the market, understanding pricing trends helps you make informed decisions about which models to use and when.
Notable Price Reductions
This month saw 11 models with measurable input price reductions. Price cuts are a strong indicator of increasing competition among providers, and they directly benefit developers and businesses relying on these APIs.
Mistral: Mistral Small 3.1 24B moved from $0.35 to $0.03 per million input tokens, a 91% reduction. Output pricing also shifted from $0.56 to $0.11/1M tokens.
MiniMax: MiniMax M2.5 moved from $0.29 to $0.12 per million input tokens, a 60% reduction. Output pricing also shifted from $1.20 to $0.99/1M tokens.
Mistral Small moved from $0.20 to $0.10 per million input tokens, a 50% reduction. Output pricing also shifted from $0.60 to $0.30/1M tokens.
Nex AGI: DeepSeek V3.1 Nex N1 moved from $0.27 to $0.14 per million input tokens, a 50% reduction. Output pricing also shifted from $1.00 to $0.50/1M tokens.
Qwen: Qwen3 Max moved from $1.20 to $0.78 per million input tokens, a 35% reduction. Output pricing also shifted from $6.00 to $3.90/1M tokens.
Qwen: Qwen3 Next 80B A3B Thinking moved from $0.15 to $0.10 per million input tokens, a 35% reduction. Output pricing also shifted from $1.20 to $0.78/1M tokens.
Qwen: Qwen VL Max moved from $0.80 to $0.52 per million input tokens, a 35% reduction. Output pricing also shifted from $3.20 to $2.08/1M tokens.
Price reductions of this magnitude often signal that a provider is either optimizing their infrastructure costs or strategically positioning a model for wider adoption. For teams currently using these models, these reductions translate directly into lower operating expenses without any code changes.
Price Increases to Watch
On the other side, 10 models saw price increases this month. While less common, price increases typically reflect added capabilities, improved quality, or changes in the underlying infrastructure costs.
- Qwen2.5 Coder 32B Instruct: $0.20 → $0.66/1M tokens (+230%)
- Google: Gemma 3 27B: $0.04 → $0.08/1M tokens (+100%)
- Mistral: Devstral Small 2505: $0.06 → $0.10/1M tokens (+67%)
- Qwen: Qwen3 30B A3B Thinking 2507: $0.05 → $0.08/1M tokens (+57%)
- Sao10K: Llama 3.1 Euryale 70B v2.2: $0.65 → $0.85/1M tokens (+31%)
If you are using any of these models, it may be worth evaluating alternatives in the same capability tier to ensure you are getting the value you need at the right price point.
New Model Releases
20 new models were added to our tracker this month. Here are the highlights:
Deepinfra
- DeepInfra: meta-llama/Meta-Llama-3.3-70B-Instruct-Turbo — unknown context, $0.12/1M input
- DeepInfra: microsoft/WizardLM-2-8x22B — unknown context, $0.48/1M input
- DeepInfra: mistralai/Mixtral-8x7B-Instruct-v0.1 — 32K context, $0.54/1M input
- DeepInfra: meta-llama/Meta-Llama-3.1-8B-Instruct — 131K context, $0.02/1M input
- DeepInfra: openchat/openchat_3.5 — unknown context, $0.06/1M input
Provider Pricing Landscape
Understanding where each provider sits in the pricing spectrum helps when evaluating options. The table below shows average, minimum, and maximum input pricing (per 1M tokens) for providers with three or more tracked models:
| Provider | Models | Avg Input | Min | Max |
|---|---|---|---|---|
| liquid | 3 | $0.02 | $0.01 | $0.03 |
| allenai | 7 | $0.14 | $0.05 | $0.20 |
| bytedance-seed | 4 | $0.17 | $0.08 | $0.25 |
| baidu | 5 | $0.20 | $0.07 | $0.42 |
| qwen | 49 | $0.24 | $0.03 | $1.04 |
| inception | 3 | $0.25 | $0.25 | $0.25 |
| minimax | 7 | $0.26 | $0.12 | $0.40 |
| tngtech | 3 | $0.28 | $0.25 | $0.30 |
| microsoft | 8 | $0.29 | $0.05 | $1.00 |
| deepseek | 16 | $0.32 | $0.06 | $0.70 |
| nvidia | 7 | $0.33 | $0.04 | $1.20 |
| groq | 21 | $0.34 | $0.05 | $1.00 |
| thedrummer | 5 | $0.43 | $0.17 | $0.75 |
| arcee-ai | 6 | $0.43 | $0.05 | $0.90 |
| nousresearch | 6 | $0.43 | $0.02 | $1.00 |
Providers with wider price ranges typically offer a lineup spanning from lightweight, cost-efficient models to premium, high-capability options. This gives developers flexibility to match model selection to specific workload requirements.
What This Means for Developers
The overall trend in AI model pricing remains downward, driven by improved hardware utilization, quantization techniques, and competitive pressure. For teams building AI-powered products, this is encouraging — the cost of intelligence continues to decrease while capabilities expand.
Practical recommendations:
- Review your model choices quarterly. A model that was cost-optimal three months ago may no longer be the right choice as new options emerge and prices shift.
- Consider multi-model strategies. Use lighter, cheaper models for simpler tasks and reserve premium models for complex reasoning or critical outputs.
- Watch for free-tier options. Several providers now offer free access to capable models with rate limits, which can significantly reduce development and testing costs.
We update pricing data daily and will continue publishing monthly trend reports. Bookmark the Trends page for the latest data between reports.