AI API Pricing Trends 2026
Every major price move across OpenAI, Anthropic, Google, DeepSeek, Mistral, Qwen, and xAI. Updated Aug 2026.
-90%
Budget model prices
since 2023
since 2023
-75%
GPT-5.4 vs GPT-4o
input price
input price
$0.03
Cheapest input
(Qwen 3.7 Flash)
(Qwen 3.7 Flash)
93
Models tracked
across 11 providers
across 11 providers
Biggest Price Moves in 2026
These are the price changes that matter most for your budget. If you're still on 2025-era models, you're overpaying.
GPT-5.4 replaced GPT-4o
OpenAI
-75%
$10.00 (GPT-4o)
→
$2.50 (GPT-5.4) / 1M input
GPT-4o deprecated. GPT-5.4 is cheaper, faster, and more capable. GPT-5.6 Luna ($0.25) is even cheaper for simpler tasks.
Qwen 3.7 Flash
Alibaba/Qwen
New!
$0.075 (prev. cheapest)
→
$0.03 / 1M input
Cheapest multimodal model available — text + image input. 1M context. Beats Gemini Flash-Lite by 60%.
Mistral Large 3
Mistral
-75%
$2.00
→
$0.50 / 1M input
Budget-tier pricing with mid-tier capabilities. Excellent value.
Claude Opus 5
Anthropic
Same price
$5.00 (Opus 4.8)
→
$5.00 (Opus 5) / 1M input
Anthropic launched Opus 5 at the same price as Opus 4.8 — more capable, no price increase. New recommended default.
Current Prices at a Glance
All prices are per 1M tokens. Sorted by tier, then by input price. Green rows = cheapest in their tier. Prices verified Aug 2026.
| Model | Provider | Tier | Input | Output | Context |
|---|---|---|---|---|---|
| Claude Opus 5 | Anthropic | Premium | $5.00 | $25.00 | 1M |
| GPT-5.5 | OpenAI | Premium | $5.00 | $30.00 | 1M |
| Claude Opus 4.1 | Anthropic | Premium | $15.00 | $75.00 | 200K |
| Grok 4.5 | xAI | Mid | $2.00 | $6.00 | 500K |
| GPT-5.4 | OpenAI | Mid | $2.50 | $15.00 | 1M |
| Claude Sonnet 5 | Anthropic | Mid | $2.00 | $10.00 | 1M |
| Gemini 3.1 Pro | Mid | $2.00 | $12.00 | 1M | |
| GPT-5 | OpenAI | Mid | $1.25 | $10.00 | 272K |
| Grok 4.3 | xAI | Mid | $1.25 | $2.50 | 1M |
| Kimi K3 | Moonshot | Mid | $2.90 | $14.00 | 1M |
| Qwen 3.7 Flash | Alibaba/Qwen | Budget | $0.03 | $0.13 | 1M |
| GPT-5 nano | OpenAI | Budget | $0.05 | $0.40 | 1M |
| Gemini 2.5 Flash-Lite | Budget | $0.075 | $0.30 | 1M | |
| GPT-oss 20B | OpenAI | Budget | $0.08 | $0.32 | 128K |
| DeepSeek V4 Flash | DeepSeek | Budget | $0.14 | $0.28 | 1M |
| Mistral Small 4 | Mistral | Budget | $0.15 | $0.60 | 128K |
| DeepSeek V4 Pro | DeepSeek | Budget | $0.44 | $0.87 | 1M |
| Mistral Large 3 | Mistral | Budget | $0.50 | $1.50 | 128K |
| Claude Haiku 4.5 | Anthropic | Budget | $1.00 | $5.00 | 200K |
Best Value Right Now
Based on current pricing (Aug 2026). These are the models that give you the most capability per dollar.
Cheapest Multimodal
Qwen 3.7 Flash
$0.03 input / $0.13 output — 1M context
Cheapest multimodal model available. Handles text + image input. 1M context window. Beats Gemini Flash-Lite by 60% on input price.
Best Budget from Major Provider
GPT-5 nano
$0.05 input / $0.40 output — 1M context
OpenAI's cheapest model. Great for classification, extraction, routing, and simple tasks. 1M context at budget pricing.
Code Generation
DeepSeek V4 Pro
$0.44 input / $0.87 output — 1M context
Strong at code completion, refactoring, and debugging. 1M context handles entire codebases. 75% cheaper than its predecessor.
Best Mid-Tier Value
GPT-5.4
$2.50 input / $15.00 output — 1M context
75% cheaper than GPT-4o with better performance. The sweet spot for most production workloads. Also try Claude Sonnet 5 ($2.00/$10.00).
Want to save these recommendations and track costs over time?
APIpulse lets you save scenarios, export cost reports, and get personalized optimization tips.
Free Tools →When to Switch Providers
Use this decision framework to decide if a provider change makes sense for your workload.
You're paying more than $0.50/1M input tokens for a non-reasoning workload
Switch to Qwen 3.7 Flash ($0.03) or DeepSeek V4 Flash ($0.14). You'll save 90%+ with similar quality for most tasks.
You're still on GPT-4o or any deprecated 2024 model
GPT-4o is deprecated. Switch to GPT-5.4 ($2.50) — same price, better quality, 1M context. For even cheaper, GPT-5 nano ($0.05) handles simple tasks at 1/50th the cost.
You need the cheapest possible API with decent quality
Qwen 3.7 Flash at $0.03/1M input is the cheapest multimodal model. GPT-5 nano ($0.05) is cheapest from OpenAI. GPT-oss 20B ($0.08) is cheapest for text-only open-weight models.
You need long context (500K+) without premium prices
Qwen 3.7 Flash offers 1M context at $0.03/$0.13. Gemini 2.5 Flash-Lite offers 1M at $0.075/$0.30. DeepSeek V4 Pro offers 1M at $0.44/$0.87. All budget-tier with generous context.
You're building a multi-model pipeline
Use Qwen 3.7 Flash for routing/classification ($0.03), DeepSeek V4 Pro for code tasks ($0.44), and Claude Sonnet 5 ($2.00) or GPT-5.4 ($2.50) for complex reasoning. Total cost under $3/1M tokens for most workloads.
How We Got Here
2023
The GPT-4 Era
GPT-4 launched at $30/$60 per 1M tokens. Claude 2 at $24. Only two serious providers. 8K-32K context windows.
Early 2024
Price Wars Begin
Google enters with Gemini. OpenAI cuts GPT-4 to $10/$30. Anthropic launches Claude 3. Context windows hit 128K.
Mid 2024
The Budget Revolution
GPT-4o mini at $0.15/$0.60. Gemini Flash at $0.075/$0.30. Open-source Llama pressures the market. Budget AI becomes viable for production.
Late 2025
Context Explodes, Prices Tank
Google launches 1M context. DeepSeek enters with aggressive pricing. Mistral Large drops 75%. GPT-4o drops 67%. 11 providers now compete.
May 2026
GPT-5 Generation Launches
OpenAI launches GPT-5 family: GPT-5, GPT-5 mini, GPT-5 nano ($0.05). Budget models hit $0.05/1M input. 1M+ context becomes standard at budget tier.
Jul 2026
Opus 5, Qwen 3.7 Flash, and 93 Models
Anthropic launches Opus 5 (same $5/$25 price as Opus 4.8). Alibaba launches Qwen 3.7 Flash at $0.03 — cheapest multimodal model. GPT-5.6 Luna/Terra/Sol replace GPT-5.4. 93 models across 11 providers. Budget input now at $0.03.
Aug 2026
Current Landscape
93 models, 11 providers. Cheapest: Qwen 3.7 Flash ($0.03). Budget floor: $0.03/1M input (was $30 in 2023 — a 99.9% drop). Mid-tier quality at budget prices is now the norm.
What to Watch Next
- Budget floor stabilizing: At $0.03/1M input, budget models are approaching cost floors. Further drops will be incremental, not dramatic
- Mid-tier price war: GPT-5.4 ($2.50), Claude Sonnet 5 ($2.00), and Gemini 3.1 Pro ($2.00) are in direct competition — expect more cuts
- Open-source pressure: Qwen, Llama, and DeepSeek open-weight models keep proprietary prices honest
- More GPT-5 variants: GPT-5.6 Luna/Terra/Sol (Jul 2026) showed OpenAI will keep launching specialized variants at different price points
Calculate your costs with current pricing. See what switching could save you.
Try the APIpulse Calculator or Compare Models Side-by-SidePrices verified against official provider pages. See full changelog for every price change. Get alerts when prices change.