AI API Price Change Tracker
Every price increase, cut, promotion, new-model price and retirement for major AI model APIs, with the price before and after, the percentage change and the official source. Prices are per 1M cache-miss input / output tokens. Updated 2026-10-25.
- 21 entries: 4 increases, 9 cuts or promotions, 4 new-model prices and 3 retirements.
- Largest increase: DeepSeek V4 Flash (2026-08-17), output $0.28 → $1.32, +371%.
- Largest cut: GPT-5.6 Luna (2026-08-01), input $1.00 → $0.2, -80%.
Coming up
K2 models stop accepting newly registered users. Existing users keep access until all K2 models retire by the end of December 2026. Source: Kimi API pricing
Search and Fetch APIs, currently free through DeepSeek Harness, Claude Code and Codex, start per-request billing at 00:00 Beijing time. Source: DeepSeek pricing
$2.00 / $10.00 → $4.00 / $20.00 (input / output, +100% / +100%)
Introductory pricing runs through November 30, 2026. Cached input rises from $0.10/M to $0.20/M. Source: Google Gemini API pricing
$0.75 / $3.75 → $1.50 / $7.50 (input / output, +100% / +100%)
The 2026 introductory price ends on December 31. Cached input also doubles to $0.15/M. Gemini 3.6 Flash follows the same schedule. Source: Google Gemini API pricing
October 2026
| Date | Model | Before | After | Change |
|---|---|---|---|---|
| 2026-10-25 first recorded | Claude Haiku 5.5 New model Anthropic Prompts above 100K tokens bill at $0.50/$2.50. Ten times cheaper than Claude Haiku 4.5 ($1/$5) for short prompts. Anthropic pricing | — | $0.1 / $0.5 | — |
| 2026-10-23 effective | Kimi K3.1 Pro New model Moonshot AI One third cheaper than Kimi K3 ($3/$15). A Highspeed endpoint costs $4/$20. Kimi API pricing | — | $2.00 / $10.00 | — |
| 2026-10-21 effective | MiMo-V2.5 and V2.5-Pro Retirement Xiaomi MiMo Retired at 10:00 Beijing time; the model IDs no longer work. MiMo-V2.6 Pro and Flash keep the same real-time token rates. Xiaomi MiMo pricing | — | — | — |
| 2026-10-19 effective | MiniMax M3.1 Flash New model MiniMax Half the input and output price of MiniMax M3 ($0.30/$1.20). MiniMax pricing | — | $0.15 / $0.6 | — |
| 2026-10-13 effective | Gemini 4 Argon New model Google Generally available with introductory pricing through November 30, 2026. Not on the free tier. Google Gemini API pricing | — | $2.00 / $10.00 | — |
September 2026
| Date | Model | Before | After | Change |
|---|---|---|---|---|
| 2026-09-25 first recorded | GPT-5.6 Sol Promotion OpenAI Promotional pricing available at least through November 21, 2026. OpenAI pricing | $5.00 / $30.00 | $4.00 / $20.00 | -20% / -33% |
| 2026-09-24 effective | Sora 2 API Retirement OpenAI All Sora API access ended. OpenAI announced the shutdown on March 24 and closed the Sora app on April 26, 2026. Sora shutdown timeline (Zilliz) | — | — | — |
| 2026-09-10 effective | DeepSeek V4.1 Flash (replaces V4 Flash) Price cut DeepSeek Peak rates shown. Off-peak fell from $0.22/$0.66 to $0.15/$0.60. V4.1 Flash also added vision input. DeepSeek pricing | $0.44 / $1.32 | $0.3 / $1.20 | -32% / -9% |
August 2026
| Date | Model | Before | After | Change |
|---|---|---|---|---|
| 2026-08-17 effective | DeepSeek V4 Flash Price increase DeepSeek Peak rates shown. DeepSeek introduced time-of-day pricing: off-peak is half the peak rate ($0.22/$0.66). DeepSeek pricing | $0.14 / $0.28 | $0.44 / $1.32 | +214% / +371% |
| 2026-08-17 effective | DeepSeek V4 Pro Price increase DeepSeek Peak rates shown. Off-peak is $0.66/$1.98. DeepSeek pricing | $0.435 / $0.87 | $1.32 / $3.96 | +203% / +355% |
| 2026-08-17 first recorded | Gemini 3.6 Flash Promotion Google Google applied the Gemini 3.7 Flash year-end introductory price to 3.6 Flash. It reverts on January 1, 2027. Google Gemini API pricing | $1.50 / $7.50 | $0.75 / $3.75 | -50% / -50% |
| 2026-08-01 first recorded | GPT-5.6 Luna Price cut OpenAI Permanent 80% reduction from launch pricing. OpenAI pricing | $1.00 / $6.00 | $0.2 / $1.20 | -80% / -80% |
| 2026-08-01 first recorded | GPT-5.6 Terra Price cut OpenAI Reduced alongside GPT-5.6 Luna at the start of August. OpenAI pricing | $2.50 / $15.00 | $2.00 / $12.00 | -20% / -20% |
| 2026-08-01 first recorded | Qwen3.7 Plus (China) Promotion Alibaba Limited-time 20% discount on the 0–256K tier in mainland China. Alibaba lists no end date. Alibaba Cloud Model Studio pricing | ¥2.00 / ¥8.00 | ¥1.60 / ¥6.40 | -20% / -20% |
May 2026
| Date | Model | Before | After | Change |
|---|---|---|---|---|
| 2026-05-31 effective | DeepSeek V4 Pro Price cut DeepSeek The launch-period 75% discount became the permanent official price after May 31, 2026 (15:59 UTC). It was later raised on August 17. DeepSeek pricing | $1.74 / $3.48 | $0.435 / $0.87 | -75% / -75% |
| 2026-05-27 effective | MiMo-V2.5-Pro Price cut Xiaomi MiMo Effective 00:00 Beijing time. The separate 256K–1M price tier was removed. Xiaomi MiMo price update | $1.00 / $3.00 | $0.435 / $0.87 | -56% / -71% |
| 2026-05-27 effective | MiMo-V2.5 Price cut Xiaomi MiMo Effective 00:00 Beijing time, together with the V2.5-Pro cut. Xiaomi MiMo price update | $0.4 / $2.00 | $0.14 / $0.28 | -65% / -86% |
DevTk.AI data corrections
These are prices DevTk.AI once recorded wrongly and later fixed. They are not provider price changes, so they are not part of the timeline above.
- 2026-10-25 · Claude Sonnet 5.5 cached input: Recorded as $0.20/M → Corrected to $0.10/M (Anthropic official pricing)
- 2026-10-25 · Qwen3.7 Max: Recorded as ¥6 / ¥18 (unverified promotion) → Corrected to ¥12 / ¥36 (official China/global list price)
- 2026-08-01 · Kimi K2.6 / K2.7 Code: Recorded as $6.50 / $27.00 (CNY values shown as USD) → Corrected to $0.95 / $4.00
- 2026-06-14 · Mistral Large 3: Recorded as $2.00 / $6.00 → Corrected to $0.50 / $1.50
- 2026-03-31 · Gemini 2.5 Flash: Recorded as $0.15 / $0.60 → Corrected to $0.30 / $2.50
- 2026-03-04 · Gemini 3.1 Pro: Recorded as $1.25 / $10.00 → Corrected to $2.00 / $12.00
- 2026-02-22 · Claude Opus 4.6: Recorded as $15 / $75 (Claude 3 Opus pricing) → Corrected to $5 / $25
- 2026-02-22 · DeepSeek V3.2: Recorded as $0.25 / $0.38 → Corrected to $0.27 / $1.10
- 2026-02-22 · Mistral Small 3.1: Recorded as $0.03 / $0.11 → Corrected to $0.20 / $0.60
- 2026-02-22 · Llama 3.3 70B: Recorded as $0.18 / $0.18 → Corrected to $0.88 / $0.88 (Together AI)
Machine-readable data
The full log is also available as JSON. You are welcome to cite it; please credit DevTk.AI: /api/price-changes.json · /api/models.json