Ling 3.1 Flash Review: A 560B Mixture-of-Experts at $0.075 per Million Tokens
Pick Ling 3.1 Flash for high-volume text work where price matters: extraction, summaries, tool steps. We haven't measured quality, so test it first.
Freedom Score
Not yet tested
- Price
- $0.113 per 1M mixed tokens (3 input : 1 output)
- Output speed
- 31 tok/s
- Success rate
- Collecting data
Ling 3.1 Flash is a hybrid-reasoning mixture-of-experts model from inclusionAI, released on 29 September 2026. The publisher describes 560B parameters in total with 25B active per token, and open weights under an MIT licence. On our API it is one OpenAI-compatible chat endpoint: up to 262,144 tokens of text in, up to 262,144 out, tools, JSON output and streaming, at $0.075 per million input tokens and $0.225 output.
- SpicyAPI Engineering
- Tested Oct 28, 2026
- Methodology v3.3 →
- We sell these models; scores ignore price →
The verdict
This is a catalogue review: we have not run our own quality or content tests. The publisher's model card lists a 1M-token window for the model itself, but the figure to plan around on this endpoint is 262,144. Benchmark scores on the model card are the publisher's own and we do not quote them.
At these prices a thousand requests that each send 10,000 tokens and get 1,000 tokens back cost about $0.98 at the listed rates, which makes it a candidate for bulk jobs. Whether the answers are good enough for your job is the open question.
Key facts
- Tested by us: catalogue check and launch smoke run only, no quality or content tests; the architecture and benchmark claims are the publisher's
- Price: $0.075 per million input tokens, $0.075 cached and $0.225 output
- Limits: 262,144-token context, replies up to 262,144 tokens (default 4,096), text-only input, up to 128 tools
- Reasoning: none, low, medium or high; medium is the default; temperature 0 to 2
- Sources: the scoring rules are on our review methodology page, and the full parameter list is on the Ling 3.1 Flash model page.
What we liked and what held it back
What we liked
- Very low price. $0.075 in and $0.225 out per million tokens.
- Thinking you can dial.
reasoning_effortruns from none to high, and at none the model answers without a thinking pass. - A large reply ceiling.
max_tokensgoes up to 262,144. - Tools, JSON and streaming on every request. Up to 128 tool definitions and
response_format. - Open weights. The publisher lists an MIT licence.
What held it back
- Text only. No image or audio parts; describe or extract first.
- Quality is unmeasured on our side. Benchmarks on the model card are the publisher's.
- Window mismatch. The model card says 1M; this endpoint takes 262,144.
- Thinking costs output tokens. Higher effort raises the cost of a reply and counts toward
max_tokens. - Release date varies by source. We use the package date, 29 September 2026.
What we tested
We have not run Ling 3.1 Flash through our capability sets or content levels. This page uses the publisher's model card, the limits and parameters of our endpoint and the live price on the model page. The publisher names agents, search, office productivity, software engineering and terminal tool use as intended uses; we have not verified any of them.
Prompts that worked
Starting points based on the endpoint's shape, not prompts we benchmarked.
1. Extract fields. A bulk job where a low price and reasoning_effort none make sense.
Extract the product name and delivery date from a short customer update.Run this prompt on Ling 3.1 Flash →
2. A tool step. One function, tool_choice set to auto.
Call report_status with status ready.Run this prompt on Ling 3.1 Flash →
Settings that matter: reasoning_effort none for extraction and classification, medium or high when the task needs working out; raise max_tokens above 4,096 when thinking is on; the publisher recommends temperature 1.0 and top_p 0.95.
Roleplay and how far it writes
We have not tested Ling 3.1 Flash on our content levels yet, so this review makes no claim about how far it writes or how often it softens or refuses. The model page carries no content-test results for it either. On our API we add no content filter; whatever comes back is the model's own behaviour.
Not tested yet: all five content levels, roleplay quality over many turns, and languages other than English. We will add results here once the runs are done.
Every character in a request should be an adult, keep real, identifiable people out of it, and you're responsible for the law where you publish.
Alternatives
- A measured low-cost model? DeepSeek V4.1 Flash has a full review and takes images as well as text.
- Faster replies on short jobs? Mercury 3 is text-only with a 1,000,000-token window.
- Images in the same call? Nemotron 3.5 Super VL accepts images and video.
How Ling 3.1 Flash earns its Spicy Index
Each dimension below links to its full workings: the table row every point came from and why points were lost.
- Public leaderboards 40% of quality
- Not yet tested
- Role-play 15% of quality
- Not yet tested
- Creative writing 15% of quality
- 64.6
- Instruction following 15% of quality
- 100.0
- Long-context memory 15% of quality
- Not yet tested
- Capability 10.5% weight
- 45.0
- Freedom Score 30% weight
- Not yet tested
No exact match on a tracked public leaderboard yet. How it works
Our role-play suite hasn’t run on this model yet. How it works
Our long-context suite hasn’t run on this model yet. How it works
Freedom evidence is untested or insufficient. The Index retains the original quality and capability score, is marked preliminary and still receives a rank. Missing evidence is not zero. How it works
Freedom evidence is untested or insufficient. The Index retains the original quality and capability score, is marked preliminary and still receives a rank. Missing evidence is not zero.
Open the full test report →Capability, item by item
Points from the public scoring table. Upscaled tiers and limits we set ourselves don’t count.
- Context window (creator-published)Not declared
- 0.0 / 30
- Input modalities on SpicyAPIText only
- 0.0 / 25
- Tool useTools and tool_choice
- 15.0 / 15
- Structured outputAccepted, schema adherence verified
- 15.0 / 15
- Reasoning control4 levels: none · low · medium · high
- 15.0 / 15
What Ling 3.1 Flash can do on SpicyAPI
Straight from the live model specification on SpicyAPI. Some creators offer more in their own apps; this is what you can call through our API today.
| Feature | Chat |
|---|---|
| Context window | 262,144 tokens |
| Max output per request | 262,144 tokens |
| Reasoning levels | none · low · medium · high |
| Tool calling | Yes |
| Structured output | Yes |
How long it really takes
Measured on real jobs over the last 30 days, from request to finished result. Times include our own test runs.
- Output speed
- 31 tok/sIncludes time to first token
- Success rate
- Collecting data
A time appears once at least 10 completed jobs are in. Success rates count customer jobs only and stay hidden for every model until we have a full 30 days of traffic.
Models to consider instead of Ling 3.1 Flash
DeepSeek
DeepSeek V4.1 Flash
A measured low-cost model?
- Spicy Index
- 81.9+5.2
- Freedom Score
- 90.4
- Price
- $0.525+367%/ 1M tokens · 3:1 input/output
Inception
Mercury 3
Faster replies on short jobs?
- Spicy Index
- 71.8−4.9
- Freedom Score
- —
- Price
- $0.325+189%/ 1M tokens · 3:1 input/output
NVIDIA
Nemotron 3.5 Super VL
Images in the same call?
- Spicy Index
- 76.8+0.1
- Freedom Score
- —
- Price
- $0.275+144%/ 1M tokens · 3:1 input/output
What do you think?
Was this review useful?
Log in to vote.
Comments
Comments are checked automatically and by our team. We remove spam, offensive content, personal attacks and anything that names a real person without consent. Report a comment and we review it within one business day.
Tried this model? Share what worked. You need a free account to comment, so we can keep spam out.
Ling 3.1 Flash FAQ
Try Ling 3.1 Flash yourself
Set every parameter in the Playground and see the exact price before you run it.
$0.113 per 1M mixed tokens (3 input : 1 output)
More reviews
See the leaderboards →GPT 6.1 Terra
Pick GPT 6.1 Terra for high-volume screens, documents and writing where reasoning_effort none helps; so far we have run only launch checks, no scored tests.
- Spicy Index
- 79.9
- Freedom Score
- —
Tested Oct 28, 2026
Read the reviewKimi K3.1 Pro
Pick Kimi K3.1 Pro for long-context work with screenshots and clips when you accept always-on thinking; so far we have run only launch checks, no scored tests.
- Spicy Index
- 80.4
- Freedom Score
- —
Tested Oct 28, 2026
Read the review
Grok 4.3
Pick Grok 4.3 for a huge context window, a genuine reasoning off switch and uncensored writing at a low price; skip it if you need file uploads.
Tested Sep 24, 2026
Read the review