MiniMax vs DeepSeek: API Cost Compared (2026)

On raw API price, DeepSeek wins at the floor: V4-Flash runs $0.14 input / $0.28 output official, while MiniMax-M2 measured via OpenRouter lands at $0.255 input / $1.00 output, so MiniMax competes on capability, not on being the cheapest token. That is the whole comparison in one line, but the asymmetry underneath it matters more than the headline rates, and it is the part most buyers skip past.

The honest framing here is that these two prices are not measured the same way, and we will say so loudly throughout. DeepSeek we can hit on its own official endpoint. MiniMax we reach through OpenRouter because there is no native key in our setup, so every MiniMax number below is the OpenRouter-routed value, not a figure pulled from MiniMax's own billing. Keep that distinction in mind; it changes how you should read the gap.

MiniMax vs DeepSeek API cost (verified 2026-06)

The cheapest-token contest is not close. DeepSeek's floor model is several times cheaper per output token than MiniMax-M2, and output is where most production bills actually concentrate.

Model	Input ($/1M)	Output ($/1M)	Source basis
DeepSeek V4-Flash	$0.14	$0.28	official api.deepseek.com
MiniMax-M2	$0.255	$1.00	measured via OpenRouter
MiniMax-M2	$0.30	$1.20	according to MiniMax pricing
MiniMax-M1	$0.40	$2.20	measured via OpenRouter

According to DeepSeek API Docs, V4-Flash bills $0.14 per million input tokens and $0.28 per million output tokens at the official endpoint, with no routing markup in between. That is the number to beat, and on a per-token basis MiniMax does not beat it. What MiniMax offers instead is a different capability tier, which is a separate purchase decision from price-per-token.

There is one wrinkle worth flagging on the MiniMax side. The OpenRouter-routed MiniMax-M2 rate we see, $0.255 / $1.00, is actually slightly cheaper than MiniMax's own listed $0.30 / $1.20. According to MiniMax, the official M2 rate is $0.30 input and $1.20 output, which we mark for native re-verification because we read it via research rather than our own MiniMax invoice. The practical upshot: route through a marketplace and you may pay under the list price, but you also inherit that marketplace's routing and availability, not MiniMax's direct SLA.

What we measured first-hand (verified 2026-06)

Documentation gives you list prices. We wanted to see what each model actually does on a real call, so we ran both, and the latency split was the surprise.

When we called MiniMax-M2 with a coding prompt through OpenRouter, the run logged 38 input tokens and 200 output tokens at a billed cost of $0.0002514, finishing in 2.95 seconds. We then hit DeepSeek V4-Flash on its official endpoint with a coding prompt and saw 27 input and 86 output tokens in 1.96 seconds, costed at the official $0.14 / $0.28 rate.

Run	Endpoint	In / Out tokens	Latency	Billed
MiniMax-M2 coding	via OpenRouter	38 / 200	2.95s	$0.0002514
DeepSeek V4-Flash coding	official api.deepseek.com	27 / 86	1.96s	computed at $0.14/$0.28

Two things stand out, and the asymmetry caveat applies to both. DeepSeek answered the coding prompt about a second faster, and it did so on its own endpoint, which is the cleaner measurement. MiniMax-M2 produced a longer completion (it ran to the 200-token output cap on the general call, and stayed verbose on coding too), which inflates its per-call cost beyond what the headline rate alone suggests. Because MiniMax came through OpenRouter and DeepSeek came direct, treat the latency gap as directional, not as a controlled head-to-head; routing adds a hop on the MiniMax side that DeepSeek's number does not carry.

Per-token gap is not the per-call gap

Here is the part buyers miss. The input rate gap looks modest, $0.14 versus $0.255, but output is where it widens: $0.28 versus $1.00 is a 3.6x difference, and output tokens dominate most chat and coding workloads. A model that also tends to write longer answers, as MiniMax-M2 did in our runs, compounds that on both axes at once.

According to DeepSeek API Docs, V4-Flash also serves repeated prompt prefixes from a server-side cache at a steep discount, which pulls its effective input cost even lower on workloads with stable system prompts. MiniMax's OpenRouter-routed rate carries no equivalent cache discount that we could verify, so on cache-friendly traffic the real-world gap is wider than the list prices imply. If your application replays a long system prompt on every call, that lever alone can decide the comparison.

DeepSeek V4-Flash is the cost floor; MiniMax-M2 is a capability play. Per output token, V4-Flash at $0.28 is roughly 3.6x cheaper than MiniMax-M2's OpenRouter-routed $1.00, and DeepSeek answered our coding prompt about a second faster on a direct call. Choose MiniMax when its model behavior earns the premium for your task, not because the token math favors it.

Definitions and the context-window caution

OpenRouter-routed pricing is a marketplace rate that an aggregator charges to relay your request to the underlying model provider, which can differ from the provider's own list price in either direction. In MiniMax-M2's case it ran slightly below MiniMax's listed rate, but it comes without the provider's direct billing relationship or SLA.

One thing we deliberately will not assert: a specific context window for MiniMax-M1. Sources conflict on it, with figures ranging from a modest window to a roughly million-token marketing claim, so we treat it as unverified. If long context is central to your decision, check the current limit in the MiniMax console rather than trusting any single number circulating online.

How we costed it

DeepSeek figures use its official published rates and a coding run against api.deepseek.com. MiniMax figures are OpenRouter-routed and live-billed; we did not hold a native MiniMax key, so the official MiniMax rates are research-sourced and flagged for native re-verification. We disclose the routing asymmetry on every comparison because it is the single most important caveat in this matchup.

FAQ

Is DeepSeek cheaper than MiniMax? On raw API price, yes. DeepSeek V4-Flash is $0.14 input / $0.28 output official, while MiniMax-M2 measured via OpenRouter is $0.255 input / $1.00 output. The output-token gap of 3.6x is the one that drives most production bills.

Why are the MiniMax and DeepSeek prices measured differently? We reach DeepSeek on its own official endpoint but reach MiniMax through OpenRouter, since we hold no native MiniMax key. Every MiniMax number here is the OpenRouter-routed value, and the official MiniMax rates are research-sourced and pending native re-verification.

Which model was faster in your test? On a coding prompt, DeepSeek V4-Flash returned in 1.96 seconds on its official endpoint versus 2.95 seconds for MiniMax-M2 via OpenRouter. Treat that as directional, because the OpenRouter hop adds latency the direct DeepSeek call does not carry.

What is MiniMax-M2's context window? We do not state a fixed number for the MiniMax-M1 lineage because sources conflict. For M2 specifically, verify the current limit in the MiniMax console before committing, rather than trusting a figure circulating online.

Should I pick MiniMax over DeepSeek? Pick MiniMax when its model behavior on your specific task earns the price premium. If you are optimizing purely for cost per token, DeepSeek V4-Flash is the cheaper floor, especially on cache-friendly workloads with stable prompts.

This is part of the MiniMax API pricing hub. For DeepSeek's full rate card, see the DeepSeek API pricing hub.

Author: Kevin Fan, Customer Success Manager at China LLM Directory, specializing in Chinese LLM ecosystem pricing. Last verified: 2026-06-26.