LongCat-2.0 vs Perceptron Mk1 vs Step 3.7 Flash
Step 3.7 Flash comes out ahead, 63 to 53 and 45 on our weighted score, and it is the cheaper option too.
Meituan (LongCat)LongCat-2.0
53/100- ECI—
- Price$0.30 / $1.20
- Context1M
Perceptron
Perceptron Mk1
45/100- ECI—
- Price$0.15 / $1.50
- Context33K
- Our pick
StepFunStep 3.7 Flash
63/100- ECI—
- Price$0.20 / $1.15
- Context256K
Step 3.7 Flash is our pick
Step 3.7 Flash is the better all-round choice, scoring 63/100 against LongCat-2.0 (53) and Perceptron Mk1 (45). It leads on price and inputs & features. LongCat-2.0 wins on context window. The score weighs price 50%, inputs & features 30%, context window 20%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.
- CapabilityNot enough dataNo independent benchmark covers every model here yet
- Lowest priceStep 3.7 FlashStep 3.7 Flash $0.438 · Perceptron Mk1 $0.487 · LongCat-2.0 $0.525 per 1M tokens (3:1 blend)
- Longest contextLongCat-2.0LongCat-2.0 1,000,000 · Step 3.7 Flash 256,000 · Perceptron Mk1 32,768 tokens
- Widest inputsPerceptron Mk1 and Step 3.7 FlashLongCat-2.0: Text · Perceptron Mk1: Text, Images, Video · Step 3.7 Flash: Text, Images, Video
- Self-hostingLongCat-2.0 and Step 3.7 FlashPublishes downloadable weights
| Measure | Weight | LongCat-2.0 | Perceptron Mk1 | Step 3.7 Flash |
|---|---|---|---|---|
| Price | 50% | 60 | 62 | 64 |
| Inputs & features | 30% | 35 | 45 | 80 |
| Context window | 20% | 60 | 0 | 36 |
| Overall | 100% | 53/100 | 45/100 | 63/100 |
Left out because at least one model lacks the data: capability. The remaining weights were rescaled.
Every spec in one table
Highlighted cells lead their row. Dashes mean the data is not published.
| Specification | LongCat-2.0Meituan (LongCat) | Perceptron Mk1Perceptron | Step 3.7 FlashStepFun |
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | — | — | — |
| ECI rank | — | — | — |
| Price per million tokens | |||
| Input | $0.30 | $0.15 (best) | $0.20 |
| Output | $1.20 | $1.50 | $1.15 (best) |
| Cached input | $0.006 (best) | — | $0.04 |
| Blended (3:1) | $0.525 | $0.487 | $0.438 (best) |
| Long-context rate | Same rate | Same rate | Same rate |
| Price source | Official LongCat API | Median of 3 providers | Official StepFun (Global) API |
| Limits | |||
| Context window | 1,000,000 tokens (best) | 32,768 tokens | 256,000 tokens |
| Max output | 131,072 tokens | 8,192 tokens | 256,000 tokens (best) |
| Inputs and features | |||
| Text | Yes | Yes | Yes |
| Images | No | Yes | Yes |
| PDFs | No | No | No |
| Audio | No | No | No |
| Video | No | Yes | Yes |
| Reasoning | Yes | Yes | Yeslow · medium · high |
| Tool calling | Yes | No | Yes |
| Structured output | No | No | Yes |
| Availability | |||
| Weights | Open | Proprietary | Open |
| API model ID | LongCat-2.0 | — | step-3.7-flash |
| API providers | 7 | 3 | 19 (best) |
| Released | Jun 30, 2026 | May 12, 2026 | May 29, 2026 |
| Knowledge cutoff | — | — | Mar 1, 2026 |
What would a month cost?
Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.
- LongCat-2.0$5.40
- Perceptron Mk1$4.50
- Step 3.7 Flash$4.30
Which should you choose?
Which is better: LongCat-2.0, Perceptron Mk1 or Step 3.7 Flash?
Step 3.7 Flash is the better all-round choice, scoring 63/100 against LongCat-2.0 (53) and Perceptron Mk1 (45). It leads on price and inputs & features. LongCat-2.0 wins on context window. The score weighs price 50%, inputs & features 30%, context window 20%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.
Which is cheaper, LongCat-2.0, Perceptron Mk1 or Step 3.7 Flash?
Step 3.7 Flash is cheaper at $0.20 input / $1.15 output per million tokens (official StepFun (Global) API price). Perceptron Mk1 costs $0.15 input / $1.50 output per million tokens (median across 3 API providers); LongCat-2.0 costs $0.30 input / $1.20 output per million tokens (official LongCat API price). At a typical mix of three input tokens to one output token, that is $0.438 per million tokens for Step 3.7 Flash versus $0.487 for Perceptron Mk1 (1.1× as much) and $0.525 for LongCat-2.0 (1.2× as much).
Which scores higher on benchmarks?
There is no independent benchmark that covers all three models yet. LongCat-2.0 has not been scored yet, Perceptron Mk1 has not been scored yet and Step 3.7 Flash has not been scored yet.
Which is better for coding?
There are no published SWE-bench Verified results for LongCat-2.0, Perceptron Mk1 and Step 3.7 Flash yet, so there is no like-for-like coding score. Test both on a sample of your own tasks. Note that Perceptron Mk1 does not support tool calling, which most coding agents need.
Which has the bigger context window?
LongCat-2.0 has the largest context window at 1,000,000 tokens, against 256,000 for Step 3.7 Flash and 32,768 for Perceptron Mk1. Maximum output per response: LongCat-2.0 up to 131,072, Perceptron Mk1 up to 8,192, Step 3.7 Flash up to 256,000 tokens.
Which can read images, PDFs, audio or video?
LongCat-2.0 accepts text; Perceptron Mk1 accepts text, images and video; Step 3.7 Flash accepts text, images and video. Perceptron Mk1 handles the widest range of inputs.
Are any of these open source?
LongCat-2.0 and Step 3.7 Flash publishes its weights and can be self-hosted; Perceptron Mk1 is proprietary.
Which is newer?
LongCat-2.0 is the newest, released Jun 30, 2026. Step 3.7 Flash came out May 29, 2026; Perceptron Mk1 came out May 12, 2026. Knowledge cutoff: Step 3.7 Flash Mar 1, 2026.
How do you decide the winner?
Each model gets a 0–100 score on capability (50%, independent benchmark results); price (25%, blended price per million tokens (3 input : 1 output), log scale); inputs & features (15%, image, PDF, audio and video input, tool calling, structured output and reasoning); context window (10%, maximum tokens per request, log scale). Dimensions missing for any model are dropped and the remaining weights rescaled, so every model is judged on the same evidence. Specs and prices come from public model listings and the labs’ own API pages; capability scores come from independent benchmark runs. Data updated Nov 2, 2026.
