tokenfactorypacketai commited on
Commit
2bbb5a5
·
verified ·
1 Parent(s): 24d898a

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +133 -1
README.md CHANGED
@@ -1,5 +1,137 @@
1
  ---
2
  license: other
3
  license_name: minimax-community
4
- license_link: LICENSE
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  license: other
3
  license_name: minimax-community
4
+ license_link: https://huggingface.co/MiniMaxAI/MiniMax-M3.1-Flash
5
+ pipeline_tag: image-text-to-text
6
+ base_model: MiniMaxAI/MiniMax-M3.1-Flash
7
+ tags:
8
+ - minimax
9
+ - minimax-m3.1
10
+ - multimodal
11
+ - mixture-of-experts
12
+ - moe
13
+ - agent
14
+ - coding
15
+ - computer-use
16
+ - long-context
17
+ - conversational
18
+ - openai-compatible
19
+ - serverless-inference
20
+ - llm-api
21
+ - token-factory
22
+ - packet-ai
23
  ---
24
+
25
+ # MiniMax-M3.1-Flash — Serverless API on Token Factory by packet.ai
26
+
27
+ Run **MiniMax-M3.1-Flash** through a **serverless, OpenAI-compatible API** with **pay-per-token** billing. [Token Factory](https://packet.ai/token-factory) by [packet.ai](https://packet.ai) serves open-weight models behind an OpenAI-compatible endpoint, so you can switch your existing OpenAI SDK code by changing only the base URL and API key.
28
+
29
+ This card describes the hosted API for [`MiniMaxAI/MiniMax-M3.1-Flash`](https://huggingface.co/MiniMaxAI/MiniMax-M3.1-Flash), MiniMax's native multimodal mixture-of-experts model — an agentic model for coding and computer use.
30
+
31
+ ## Why use MiniMax-M3.1-Flash on Token Factory
32
+
33
+ - **OpenAI-compatible endpoint** — works with existing OpenAI SDKs.
34
+ - **Streaming, JSON mode and tool calling** supported on chat completions.
35
+ - **Pay per token** — input and output tokens are metered separately; nothing is charged while idle.
36
+ - **Free for one month at launch (Dec 16th)**, then billed per 1M tokens. See [current pricing](https://packet.ai/token-factory).
37
+ - **Scale to dedicated GPUs** — move the model to a dedicated GPU running vLLM while keeping the same endpoint shape.
38
+
39
+ ## Good for
40
+
41
+ MiniMax describes M3.1-Flash as built for:
42
+
43
+ - **Coding**
44
+ - **Agentic workflows**
45
+ - **Computer use**
46
+ - **Cybersecurity**
47
+
48
+ ## Model overview
49
+
50
+ | Property | Value |
51
+ |---|---|
52
+ | Model | MiniMax-M3.1-Flash |
53
+ | Developer | MiniMax |
54
+ | Type | Native multimodal mixture-of-experts (MoE) |
55
+ | Builds on | MiniMax-M3 |
56
+ | Attention | MiniMax Sparse Attention (MSA) |
57
+ | Context window | Up to 1M tokens (usable length depends on serving configuration) |
58
+ | Input | Text, image, video |
59
+ | Output | Text |
60
+ | Reasoning effort levels | low, medium, high, xhigh, max |
61
+ | Recommended sampling | temperature = 1.0, top_p = 0.95 |
62
+ | License | MiniMax Community License |
63
+
64
+ ## Benchmarks
65
+
66
+ Selected results reported by MiniMax for the official MiniMax-M3.1-Flash weights. See the [original model card](https://huggingface.co/MiniMaxAI/MiniMax-M3.1-Flash) for the full table and comparisons.
67
+
68
+ | Benchmark | MiniMax-M3.1-Flash |
69
+ |---|---|
70
+ | DeepSWE v1.1 | 73.7 |
71
+ | NL2Repo-Bench | 61.2 |
72
+ | Terminal Bench 4.0 | 38.1 |
73
+ | OSWorld 2.0 | 61.7 |
74
+ | Automation-Bench v1.0.6 | 62.8 |
75
+ | BankerToolBench | 76.2 |
76
+ | CyberGym | 82.2 |
77
+ | LOCA-Bench 256K | 60.3 |
78
+ | BabyVision (with tools) | 91.0 |
79
+ | CharXiv RQ | 82.5 |
80
+ | HLE (without tools) | 43.4 |
81
+
82
+ ## Quickstart
83
+
84
+ ### 1. Get an API key
85
+
86
+ Sign up at [dash.packet.ai](https://dash.packet.ai/signup) and create an API key.
87
+
88
+ ### 2. Call the API (Python, OpenAI SDK)
89
+
90
+ ```python
91
+ from openai import OpenAI
92
+
93
+ client = OpenAI(
94
+ api_key="PACKET_API_KEY",
95
+ base_url="https://api.packet.ai/v1",
96
+ )
97
+
98
+ response = client.chat.completions.create(
99
+ model="MINIMAX_MODEL_ID_FROM_DASHBOARD",
100
+ messages=[{"role": "user", "content": "Implement Hello World"}],
101
+ )
102
+ print(response.choices[0].message.content)
103
+ ```
104
+
105
+ Replace `PACKET_API_KEY` with your API key. Full API documentation: [dash.packet.ai/docs](https://dash.packet.ai/docs)
106
+
107
+ ## Other models on Token Factory
108
+
109
+ | Model | Highlights |
110
+ |---|---|
111
+ | [`nvidia/DeepSeek-V4.1-Flash-NVFP4`](https://huggingface.co/nvidia/DeepSeek-V4.1-Flash-NVFP4) | Reasoning model, 1M-token context, image input |
112
+ | [`Qwen/Qwen2.5-1.5B-Instruct`](https://huggingface.co/tokenfactorypacketai/Qwen2.5-1.5B-Instruct) | Small, fast model for classification and extraction, 32K context, 29+ languages |
113
+
114
+ See the full, up-to-date list at [packet.ai/token-factory](https://packet.ai/token-factory).
115
+
116
+ ## Run it yourself
117
+
118
+ Prefer to self-host? Use the original weights from MiniMax: [`MiniMaxAI/MiniMax-M3.1-Flash`](https://huggingface.co/MiniMaxAI/MiniMax-M3.1-Flash). MiniMax recommends SGLang, vLLM or TokenSpeed for serving.
119
+
120
+ ## About packet.ai
121
+
122
+ [packet.ai](https://packet.ai) is a hosted·ai project operated by HOSTED AI Inc., 2880 Zanker Rd, Suite 203, San Jose, CA 95134. Alongside Token Factory, packet.ai offers GPUs (B200, RTX PRO 6000, A100, RTX 4090), dynamic and dedicated GPUs, and bare metal/clusters, billed by the minute.
123
+
124
+ - Website: [packet.ai](https://packet.ai)
125
+ - Token Factory: [packet.ai/token-factory](https://packet.ai/token-factory)
126
+ - Docs: [dash.packet.ai/docs](https://dash.packet.ai/docs)
127
+ - Talk to an engineer: [packet.ai/contact](https://packet.ai/contact)
128
+ - X: [@packet_ai](https://x.com/packet_ai) · LinkedIn: [packet-dot-ai](https://www.linkedin.com/company/packet-dot-ai) · GitHub: [hosted-ai](https://github.com/hosted-ai)
129
+
130
+ ## License and attribution
131
+
132
+ MiniMax-M3.1-Flash is developed by [MiniMax](https://huggingface.co/MiniMaxAI) and released under the MiniMax Community License; see the [original model card](https://huggingface.co/MiniMaxAI/MiniMax-M3.1-Flash) for the license terms. Token Factory provides hosted inference for this model. Use of the API is subject to packet.ai's [Terms](https://packet.ai/terms) and [Privacy Policy](https://packet.ai/privacy).
133
+
134
+ ## Citation
135
+
136
+ MiniMax Sparse Attention paper: [arXiv:2606.13392](https://arxiv.org/abs/2606.13392)
137
+