DeepSeek-V4.1-Flash is the stronger coder with the cleaner licence; MiniMax-M3.1-Flash is the easier one to self-host and reads video, but hallucinates more readily.
Choose DeepSeek-V4.1-Flash if coding agents are the main job, you want an MIT licence with no revenue cap, and you can schedule work off-peak or use a third-party host.
Choose MiniMax-M3.1-Flash if you need video input, a flat price at any hour, or weights that fit on one eight-GPU node, and you can check its answers before they reach anyone.
DeepSeek-V4.1-Flash vs MiniMax-M3.1-Flash, side by side
Feature
DeepSeek-V4.1-Flash
MiniMax-M3.1-Flash
Our rating
4.2/5
4.0/5
Price
Free MIT-licensed weights. DeepSeek API: $0.15 input / $0.60 output per 1M tokens off-peak, $0.30 / $1.20 at weekday peak; cache hits $0.003 to $0.006.
Open weights under the MiniMax community licence. MiniMax API per 1M tokens, flat at any prompt length: $0.15 input, $0.60 output, $0.015 cached input.
Best for
Cost-sensitive teams running coding and automation agents who can use a China-hosted API or a third-party host, or have multi-GPU servers for the open weights.
Teams running high-volume agent and automation work, or long multimodal inputs, who can verify outputs and accept MiniMax's licence terms.
API price per 1M tokens
$0.15 in / $0.60 out off-peak; $0.30 / $1.20 at weekday peak
$0.15 in / $0.60 out at any hour and prompt length
Cached input
$0.003 to $0.006
$0.015
Size and checkpoint
763B total, 8B to 16B active; about 765 GB
About 428B total, 23B active; about 250 GB in MXFP8
Inputs and output cap
Text and images; 384K output
Text, images and video; 524,288 output on OpenRouter
Reasoning control
Effort from 1 to 100; thinking on by default
Five levels, low to max; max by default; cannot be disabled
Licence
MIT, no use restrictions
MiniMax community licence: credit, notice, permission above $20M revenue
Artificial Analysis Intelligence Index
45 for the 1017 release, about $0.26 per task
43, about $0.22 per task
Coding (self-reported, MiniMax's table)
77.0% DeepSWE v1.1; 42.3 Terminal Bench 4.0
73.7% DeepSWE v1.1; 38.1 Terminal Bench 4.0
API data terms
Stored in China; may be used for training, opt-out on request
Retention and training terms not published on pages we could read
Both are open-weight mixture-of-experts models with a 1M-token context at $0.15 input and $0.60 output per million tokens on their makers' APIs. DeepSeek charges that only off-peak and doubles it in weekday peak hours; MiniMax charges it flat. DeepSeek leads on coding benchmarks and carries an MIT licence, while MiniMax ships a checkpoint a third the size under a licence with attribution and revenue conditions.
On paper these two cost the same, and in practice the choice turns on timing, licence and how much you trust the answers. DeepSeek's API charges $0.15 per million input tokens and $0.60 per million output only off-peak; between 01:00 and 04:00 and 06:00 and 10:00 UTC on weekdays both double. MiniMax charges $0.15 and $0.60 flat, at any hour and any prompt length. Cached input tips the other way: DeepSeek's cache hits cost $0.003 to $0.006 per million, MiniMax's $0.015, so a long system prompt resent thousands of times is cheaper on DeepSeek.
DeepSeek has the better scores where most teams will use these models. Artificial Analysis rates the 1017 release at 45 on its Intelligence Index to M3.1-Flash's 43, according to VentureBeat, and MiniMax's own comparison table shows DeepSeek ahead on DeepSWE v1.1 (77.0% to 73.7%), Terminal Bench 4.0 (42.3 to 38.1) and NL2Repo-Bench (68.2 to 61.2). MiniMax leads on Automation-Bench v1.0.6, 62.8 to 60.5. The card figures are all self-reported. The independent warning is on MiniMax's side: VentureBeat reports a 93% hallucination rate for M3.1-Flash on AA-Omniscience, meaning that when it was not right it nearly always answered wrongly instead of declining.
Reasoning costs are where the sticker prices mislead. M3.1-Flash always reasons and defaults to its maximum effort, which MiniMax's docs say spends more output tokens; Artificial Analysis calls it very verbose, at 180 million output tokens to run its index against a median of 100 million. Teams on MiniMax should set effort per request rather than accept the default. DeepSeek exposes effort on a 1 to 100 scale, with thinking on by default.
For self-hosting the balance flips. DeepSeek's MIT licence has no conditions, but its checkpoint is about 765 GB and overflows an eight-way H100 80 GB node. MiniMax's is about 250 GB and fits on one, yet its licence asks for a "Built with MiniMax M3.1-Flash" credit, a notice email, and written permission for products above $20 million in yearly revenue. For a large company that wants weights on its own servers, DeepSeek's paperwork is lighter and its hardware bill heavier.
Which should you choose?
The jobs each one is best at, from our reviews.
Pick DeepSeek-V4.1-Flash if you are…
A developer-tools startup running background coding agents across 2,000 repositories
Resolve routine issues and dependency bumps overnight
Schedule runs outside 01:00 to 04:00 and 06:00 to 10:00 UTC to pay the off-peak $0.15/$0.60 rate; DeepSeek reports 77.0% on DeepSWE v1.1 with the mini-SWE harness. Expect less in Claude Code or Codex harnesses, which scored 4 to 9 points lower on the September card.
An operations team automating back-office work across SaaS tools
Run long multi-step automations that involve screenshots
Image input and a 1M-token context let one session hold tool output and screenshots, and the card's 60.5% on AutomationBench tops its comparison table. The API's 2,500-request concurrency limit rarely caps a mid-size deployment.
A CTO at a European fintech
Decide whether DeepSeek's API is acceptable for customer data
DeepSeek's privacy policy places processing and storage in China and lists model training among its uses, so customer data belongs on a self-hosted or EU-hosted deployment instead. The MIT licence allows that with no revenue cap, but the weights alone need more than 765 GB of GPU memory, for example eight 141 GB H200s.
Pick MiniMax-M3.1-Flash if you are…
An operations team automating back-office work across a dozen SaaS tools
Run multi-step automations that read screenshots and fill in forms
MiniMax reports 62.8 on Automation-Bench v1.0.6 and 61.7 on OSWorld 2.0, both self-reported. At $0.15 input and $0.60 output, 10 billion input and 1 billion output tokens a month comes to about $2,100 before cache hits, and setting effort to low or medium on routine steps cuts the reasoning tokens billed as output.
A legal-tech engineer working on M&A data rooms
Ask questions across a whole contract bundle in one prompt
Price does not rise past 512K tokens as M3's does, so a full 1M-token prompt costs $0.15 in input, a quarter of M3's long-context rate, and $0.015 on cached re-reads. Given the hallucination rate, every answer should quote the clause it relies on so a reviewer can check it.
A support lead at a 200-person software company
Draft replies to customer tickets from the help centre
1M replies a month at 2,000 input and 300 visible output tokens is about $480, plus the reasoning tokens that the default max effort adds on top. With a 93% hallucination rate on questions it gets wrong, drafts need retrieval from the help centre and a human send button rather than auto-replies.
DeepSeek-V4.1-Flash vs MiniMax-M3.1-Flash pricing
DeepSeek-V4.1-Flash
DeepSeek-V4.1-Flash plans
Weights
FreeMIT licence
API off-peak
$0.15 in / $0.60 out per 1M tokens
API weekday peak
$0.30 in / $1.20 out per 1M tokens
Cache hits
$0.003 to $0.006 per 1M tokens
MiniMax-M3.1-Flash
Open weights under the MiniMax community licence. MiniMax API per 1M tokens, flat at any prompt length: $0.15 input, $0.60 output, $0.015 cached input.
Pros and cons
DeepSeek-V4.1-Flash
Good
MIT licence with image input
$0.15 / $0.60 per 1M tokens off-peak; cache hits from $0.003
77.0% on DeepSWE v1.1 (self-reported, 1017)
KV cache of 890 bytes per token keeps 1M contexts cheap to hold
Not so good
Data stored in China; API training terms not spelled out
Trails Opus 5 on Terminal-Bench 4.0 and HLE
About 765 GB checkpoint for self-hosting
Benchmarks self-reported and harness-sensitive
MiniMax-M3.1-Flash
Good
$0.15 / $0.60 per 1M tokens at any prompt length
Intelligence Index 43, fifth of 173 in its class
Text, image and video input with 1M context
About 250 GB checkpoint fits one 8x H100 node
Not so good
93% hallucination rate on AA-Omniscience, per VentureBeat
Reasoning cannot be disabled; default max effort is verbose
Licence needs attribution, a notice and permission above $20M revenue
Trails DeepSeek-V4.1-Flash-1017 on coding benchmarks (self-reported)
Before you commit
DeepSeek-V4.1-Flash vs MiniMax-M3.1-Flash: setup, integrations, data, cost and limits
Topic
DeepSeek-V4.1-Flash
MiniMax-M3.1-Flash
Setup
OpenAI- or Anthropic-format API at api.deepseek.com with model deepseek-flash; thinking mode is on by default. For local use there is no Jinja chat template: DeepSeek ships a Python encoder and the deepseek-recipe Rust library instead.
API: create a key on platform.minimax.io and call MiniMax-M3.1-Flash through the Anthropic-compatible endpoint (recommended) or the OpenAI-compatible one, setting reasoning_effort per request. Weights: hf download MiniMaxAI/MiniMax-M3.1-Flash and serve with SGLang, vLLM or TokenSpeed; MiniMax recommends temperature 1.0 and top_p 0.95.
Integrations
JSON output, tool calls, the Responses API, an Anthropic-compatible API and vision on DeepSeek's API; documented setups for Claude Code, Codex, OpenCode and DeepSeek Harness. OpenRouter lists third-party hosts including DeepInfra and Baidu. Search (POST /v1/tools/search, 1 to 50 results, site: and date operators) and Fetch (POST /v1/tools/fetch, Markdown output) use the same API key, and the three agent harnesses call them automatically as server-side tools. For self-hosting, DeepSeek's Apache 2.0 DeepInfer engine supports this model on NVIDIA SM100 and Ascend 950, offline only.
MiniMax API, MiniMax Code and the M Plan subscription; on OpenRouter at the same $0.15/$0.60 from MiniMax and about a dozen hosts including Together, Nebius and Novita. Transformers, KTransformers and Unsloth support is listed as in progress.
Data and privacy
Weights under MIT with no use restrictions. On DeepSeek's service, the privacy policy (updated 10 February 2026) says data is stored in China, kept while the account exists and may train models, with an opt-out on request; no separate API retention terms are published.
Licence: commercial use needs a "Built with MiniMax M3.1-Flash" credit and a one-time notice to MiniMax; products over $20 million in yearly revenue need prior written authorisation; military use is prohibited. MiniMax's API retention and training terms were not published on the pages we could read.
Cost at scale
10 billion input and 1 billion output tokens a month costs about $2,100 off-peak or $4,200 at peak on DeepSeek; cached input is $0.003 per million off-peak. Search and Fetch add $0.75 per 1,000 successful calls each, so an agent making 20 searches per task across 10,000 tasks pays $150 a month for search; free through harnesses until 31 October 2026, 16:00 UTC. Self-hosting needs over 765 GB of GPU memory for the weights alone, and DeepInfer's deployment notes put DeepSeek's smallest efficient unit at 64 prefill plus 64 decode GPUs.
10 billion input and 1 billion output tokens a month is about $2,100 on MiniMax's API, with cached input at $0.015 per million; reasoning at max effort inflates the output side. Self-hosting needs about 250 GB of GPU memory for the MXFP8 weights before KV cache, within an eight-GPU H100 80 GB node.
Limits
1M-token context and 384K output; 2,500 concurrent requests on the API. Legacy deepseek-v4-flash names now route to the 1017 model. Search is capped at 100 queries per second and Fetch at 25, per account across all keys, with a per-domain limit on Fetch and HTTP 429 beyond it.
1M-token context; OpenRouter lists 524,288 output tokens on MiniMax's own endpoint. Thinking cannot be disabled, and a request that tries returns a 400 error. MiniMax publishes no rate limits for the model, and Artificial Analysis's listing still described the weights as unreleased when we checked.
Common questions
Is DeepSeek-V4.1-Flash better than MiniMax-M3.1-Flash?
DeepSeek-V4.1-Flash is the stronger coder with the cleaner licence; MiniMax-M3.1-Flash is the easier one to self-host and reads video, but hallucinates more readily. We rate DeepSeek-V4.1-Flash 4.2 and MiniMax-M3.1-Flash 4.0 out of 5.
When should I choose DeepSeek-V4.1-Flash?
If coding agents are the main job, you want an MIT licence with no revenue cap, and you can schedule work off-peak or use a third-party host.
When should I choose MiniMax-M3.1-Flash?
If you need video input, a flat price at any hour, or weights that fit on one eight-GPU node, and you can check its answers before they reach anyone.
How much do DeepSeek-V4.1-Flash and MiniMax-M3.1-Flash cost?
DeepSeek-V4.1-Flash: Free MIT-licensed weights. DeepSeek API: $0.15 input / $0.60 output per 1M tokens off-peak, $0.30 / $1.20 at weekday peak; cache hits $0.003 to $0.006. MiniMax-M3.1-Flash: Open weights under the MiniMax community licence. MiniMax API per 1M tokens, flat at any prompt length: $0.15 input, $0.60 output, $0.015 cached input.
