Beam 1.0 501B A23B vs LongCat-2.5-Preview vs Perceptron Mk1.5
LongCat-2.5-Preview comes out ahead, 60 to 55 and 54 on our weighted score, though Beam 1.0 501B A23B is 10% cheaper per token.
Reflection AIBeam 1.0 501B A23B
54/100- ECI—
- Price$0.30 / $1.00
- Context1.05M
- Our pick
Meituan (LongCat)LongCat-2.5-Preview
60/100- ECI—
- Price$0.30 / $1.20
- Context1M
Perceptron
Perceptron Mk1.5
55/100- ECI—
- Price$0.154 / $1.54
- Context37K
LongCat-2.5-Preview is our pick
LongCat-2.5-Preview is the better all-round choice, scoring 60/100 against Perceptron Mk1.5 (55) and Beam 1.0 501B A23B (54). Perceptron Mk1.5 wins on inputs & features. 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 priceBeam 1.0 501B A23BBeam 1.0 501B A23B $0.475 · Perceptron Mk1.5 $0.502 · LongCat-2.5-Preview $0.525 per 1M tokens (3:1 blend)
- Longest contextBeam 1.0 501B A23BBeam 1.0 501B A23B 1,048,576 · LongCat-2.5-Preview 1,000,000 · Perceptron Mk1.5 36,864 tokens
- Widest inputsPerceptron Mk1.5Beam 1.0 501B A23B: Text · LongCat-2.5-Preview: Text, Images · Perceptron Mk1.5: Text, Images, Audio, Video
- Self-hostingBeam 1.0 501B A23BPublishes downloadable weights (Apache-2.0)
| Measure | Weight | Beam 1.0 501B A23B | LongCat-2.5-Preview | Perceptron Mk1.5 |
|---|---|---|---|---|
| Price | 50% | 62 | 60 | 61 |
| Inputs & features | 30% | 35 | 60 | 80 |
| Context window | 20% | 61 | 60 | 2 |
| Overall | 100% | 54/100 | 60/100 | 55/100 |
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 | Beam 1.0 501B A23BReflection AI | LongCat-2.5-PreviewMeituan (LongCat) | Perceptron Mk1.5Perceptron |
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | — | — | — |
| ECI rank | — | — | — |
| Price per million tokens | |||
| Input | $0.30 | $0.30 | $0.154 (best) |
| Output | $1.00 (best) | $1.20 | $1.54 |
| Cached input | — | $0.006 | — |
| Blended (3:1) | $0.475 (best) | $0.525 | $0.502 |
| Long-context rate | Same rate | Same rate | Same rate |
| Price source | Median of 7 providers | Official LongCat API | Median of 2 providers |
| Limits | |||
| Context window | 1,048,576 tokens (best) | 1,000,000 tokens | 36,864 tokens |
| Max output | 131,072 tokens (best) | 131,072 tokens (best) | 8,192 tokens |
| Inputs and features | |||
| Text | Yes | Yes | Yes |
| Images | No | Yes | Yes |
| PDFs | No | No | No |
| Audio | No | No | Yes |
| Video | No | No | Yes |
| Reasoning | Yes | Yes | Yes |
| Tool calling | Yes | Yes | Yes |
| Structured output | No | No | No |
| Availability | |||
| Weights | OpenApache-2.0 | Proprietary | Proprietary |
| API model ID | — | LongCat-2.5-Preview | — |
| API providers | 12 (best) | 4 | 2 |
| Released | Oct 5, 2026 | Sep 25, 2026 | Sep 25, 2026 |
| Knowledge cutoff | Jun 30, 2026 | — | — |
What would a month cost?
Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.
- Beam 1.0 501B A23B$5.00
- LongCat-2.5-Preview$5.40
- Perceptron Mk1.5$4.63
Which should you choose?
Which is better: Beam 1.0 501B A23B, LongCat-2.5-Preview or Perceptron Mk1.5?
LongCat-2.5-Preview is the better all-round choice, scoring 60/100 against Perceptron Mk1.5 (55) and Beam 1.0 501B A23B (54). Perceptron Mk1.5 wins on inputs & features. 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, Beam 1.0 501B A23B, LongCat-2.5-Preview or Perceptron Mk1.5?
Beam 1.0 501B A23B is cheaper at $0.30 input / $1.00 output per million tokens (median across 7 API providers). Perceptron Mk1.5 costs $0.154 input / $1.54 output per million tokens (median across 2 API providers); LongCat-2.5-Preview 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 $0.475 per million tokens for Beam 1.0 501B A23B versus $0.502 for Perceptron Mk1.5 (1.1× as much) and $0.525 for LongCat-2.5-Preview (1.1× as much).
Which scores higher on benchmarks?
There is no independent benchmark that covers all three models yet. Beam 1.0 501B A23B has not been scored yet, LongCat-2.5-Preview has not been scored yet and Perceptron Mk1.5 has not been scored yet.
Which is better for coding?
There are no published SWE-bench Verified results for Beam 1.0 501B A23B, LongCat-2.5-Preview and Perceptron Mk1.5 yet, so there is no like-for-like coding score. Test both on a sample of your own tasks. All three support tool calling for agent workflows.
Which has the bigger context window?
Beam 1.0 501B A23B has the largest context window at 1,048,576 tokens, against 1,000,000 for LongCat-2.5-Preview and 36,864 for Perceptron Mk1.5. Maximum output per response: Beam 1.0 501B A23B up to 131,072, LongCat-2.5-Preview up to 131,072, Perceptron Mk1.5 up to 8,192 tokens.
Which can read images, PDFs, audio or video?
Beam 1.0 501B A23B accepts text; LongCat-2.5-Preview accepts text and images; Perceptron Mk1.5 accepts text, images, audio and video. Perceptron Mk1.5 handles the widest range of inputs.
Are any of these open source?
Beam 1.0 501B A23B publishes its weights (Apache-2.0) and can be self-hosted; LongCat-2.5-Preview and Perceptron Mk1.5 is proprietary.
Which is newer?
Beam 1.0 501B A23B is the newest, released Oct 5, 2026. LongCat-2.5-Preview came out Sep 25, 2026; Perceptron Mk1.5 came out Sep 25, 2026. Knowledge cutoff: Beam 1.0 501B A23B Jun 30, 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.
