Lower Prices for DeepSeek V4.1 Flash and MiniMax H3 (November 17, 2026)
DeepSeek V4.1 Flash at Tier 3 now costs 55% of the official price, and MiniMax H3 costs 70%, 65% and 50% of the official price at Tiers 1, 2 and 3. Here is what each tier pays.
By OceanToken Team
Short answer: From November 17, 2026, DeepSeek V4.1 Flash at Tier 3 costs 55% of the official price, so you save up to 45%. MiniMax H3 costs 70%, 65% and 50% of the official price at Tiers 1, 2 and 3, so you save up to 50%. You do not need to change anything: the new prices apply to every request from today.
DeepSeek V4.1 Flash
Prices per 1M tokens, input / output.
| Price | Off the official price | |
|---|---|---|
| Official price | $0.30 / $1.20 | |
| Tier 1 | $0.21 / $0.84 | 30% |
| Tier 2 | $0.195 / $0.78 | 35% |
| Tier 3 | $0.165 / $0.66 | 45% (was 40%) |
Tier 1 and Tier 2 are unchanged. Cached input is billed at the official cache price at every tier.
MiniMax H3
Prices per second of video.
| 768p | 2K | Off the official price | |
|---|---|---|---|
| Official price | $0.08 | $0.13 | |
| Tier 1 | $0.056 | $0.091 | 30% |
| Tier 2 | $0.052 | $0.0845 | 35% |
| Tier 3 | $0.04 | $0.065 | 50% |
A 5-second 768p clip costs $0.28 at Tier 1 and $0.20 at Tier 3, against $0.40 at the official price.
Which tier you are in
Your tier follows your total top-ups: Tier 1 below $1,000, Tier 2 from $1,000 and Tier 3 from $10,000. Your console shows the tier and the rate that apply to you. See How tiers work for the details, and the DeepSeek V4.1 Flash and MiniMax H3 pages for every price.
Spending more than $1,000 a month on one model? Send us your bill and we will quote a price for your volume.
More from the blog
8 Models Retired on November 16, 2026, and What to Use Instead
Seed 2.0 Code, Seed 1.6, Seed 1.6 Flash, Qwen 2.5 7B, Qwen3 VL 30B Thinking, Hunyuan A13B, Step 3.5 Flash and LFM 2.5 2.6B (free) have left the catalog. Here is the model to switch to for each one.
o3-pro Retired on November 14, 2026, and What to Use Instead
OpenAI's o3-pro is no longer available on OceanToken from November 14, 2026. Change the model field to openai/gpt-5-pro or openai/gpt-5.5-pro and your requests work as before.
29 Models Retired on November 10, 2026, and What to Use Instead
Their makers stopped serving 29 models, including o1, o3-mini, o4-mini, GPT-4 Turbo, GPT-3.5 Turbo and MiniMax-M1. Here is the model to switch to for each one.