MiniMax
Chat with MiniMax M3 online
A million tokens in, a very long answer out, with vision — MiniMax's multimodal model for long-horizon work.
1.048576M tokens
Context
262,144 tokens
Max output
Fast
Speed
Straight from MiniMax M3
Three prompts, three unedited answers, generated on September 2, 2026 through NinjaChat. Click the prompt to copy it and run it yourself.
Write a Python function that merges overlapping intervals, plus three short tests. Keep the whole answer under 40 lines of code.
def merge_intervals(intervals):
if not intervals:
return []
intervals = sorted(intervals, key=lambda x: x[0])
merged = [intervals[0][:]
for start, end in intervals[1:]:
if start <= merged[-1][1]:
merged[-1][1] = max(merged[-1][1], end)
else:
merged.append([start, end])
return merged
# Tests
assert merge_intervals([[1, 3], [2, 6], [8, 10], [15, 18]]) == [[1, 6], [8, 10], [15, 18]]
assert merge_intervals([[1, 4], [4, 5]]) == [[1, 5]]
assert merge_intervals([]) == []
assert merge_intervals([[1, 4], [2, 3], [5, 7]]) == [[1, 4], [5, 7]]
print("All tests passed!")
Read: We ran the same 3 prompts on 12 AI models — Unedited answers, latency and price for every model, measured through the NinjaChat API in September 2026.
Served by Fireworks as minimax-m3. Answers vary run to run; these are one take each, not the best of several.
What you’ll use it for
Long documents and recordings
A million tokens in, a very long answer out.
Video and image understanding
Frames, screenshots and diagrams as input.
Coding walkthroughs
Read a repository end to end and explain or rewrite it.
Long-form writing
Reports, specs and drafts that run long.
Prompts to steal
Tuned to this model — click any line to copy.
About MiniMax M3
MiniMax M3 is the successor to the M2 line, released on May 31, 2026: a multimodal foundation model that reads images and video, holds a million tokens of context, and can write up to 262K tokens in one answer. MiniMax built it for long-horizon agentic work and coding, and used a sparse-attention design that keeps the cost of all that context down. On NinjaChat it is included in every plan.
The combination is unusual in its class: very long input, very long output, and vision, at a price that makes it practical to use for the whole of a big job rather than a slice. That suits document-heavy work, a long recording you want turned into notes and actions, or a repository you want walked end to end.
One limit to know: on NinjaChat's rail, M3 does not accept function calling, so it is the model for reading, reasoning and writing rather than for driving tools. When a task needs tool calls, Kimi K3 or GLM 5.2 are the open-model picks beside it, and Claude and GPT-5.6 sit in the same picker.
How to use MiniMax M3
From zero to your first result in under a minute.
01
Create a NinjaChat account and choose a plan
02
Open chat and pick MiniMax M3 from the model list
03
Attach the documents, images or video and ask
04
Ask for the full output; it can write very long answers
05
Switch to Kimi K3 or GLM 5.2 when you need tool calls
Tips for better results
- Use it for reading and writing at length
- Give it the recording or the screenshot directly
- Keep the whole project in one thread
- Escalate to a frontier model for the hardest reasoning
MiniMax M3 vs the alternatives
Where it wins, and where it doesn't.
Questions, answered.
Beyond chat — images & video
The same subscription covers the image studio and video generation.
MiniMax M3, and 50 more.
One subscription covers every model on NinjaChat, MiniMax M3 included.
