Step 3.5 Flash 2603 vs Gemma 4 12B IT vs MiMo-V2.5
MiMo-V2.5 comes out ahead, 79 to 75 and 62 on our weighted score, though Gemma 4 12B IT is 29% cheaper per token.
StepFunStep 3.5 Flash 2603
62/100- ECI—
- Price$0.10 / $0.30
- Context256K
GoogleGemma 4 12B IT
75/100- ECI—
- Price$0.075 / $0.275
- Context262K
- Our pick
XiaomiMiMo-V2.5
79/100- ECI—
- Price$0.14 / $0.28
- Context1.05M
MiMo-V2.5 is our pick
MiMo-V2.5 is the better all-round choice, scoring 79/100 against Gemma 4 12B IT (75) and Step 3.5 Flash 2603 (62). It leads on inputs & features and context window. Gemma 4 12B IT wins on price. 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 priceGemma 4 12B ITGemma 4 12B IT $0.125 · Step 3.5 Flash 2603 $0.15 · MiMo-V2.5 $0.175 per 1M tokens (3:1 blend)
- Longest contextMiMo-V2.5MiMo-V2.5 1,048,576 · Gemma 4 12B IT 262,144 · Step 3.5 Flash 2603 256,000 tokens
- Widest inputsMiMo-V2.5Step 3.5 Flash 2603: Text · Gemma 4 12B IT: Text, Images · MiMo-V2.5: Text, Images, Audio, Video
- Self-hostingAll open weightsEvery model here can be downloaded and run on your own hardware
| Measure | Weight | Step 3.5 Flash 2603 | Gemma 4 12B IT | MiMo-V2.5 |
|---|---|---|---|---|
| Price | 50% | 89 | 93 | 86 |
| Inputs & features | 30% | 35 | 70 | 80 |
| Context window | 20% | 36 | 37 | 61 |
| Overall | 100% | 62/100 | 75/100 | 79/100 |
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 | Step 3.5 Flash 2603StepFun | Gemma 4 12B ITGoogle | MiMo-V2.5Xiaomi |
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | — | — | — |
| ECI rank | — | — | — |
| Price per million tokens | |||
| Input | $0.10 | $0.075 (best) | $0.14 |
| Output | $0.30 | $0.275 (best) | $0.28 |
| Cached input | $0.02 | — | $0.0028 (best) |
| Blended (3:1) | $0.15 | $0.125 (best) | $0.175 |
| Long-context rate | Same rate | Same rate | Same rate |
| Price source | Official StepFun (Global) API | Median of 2 providers | Official Xiaomi API |
| Limits | |||
| Context window | 256,000 tokens | 262,144 tokens | 1,048,576 tokens (best) |
| Max output | 256,000 tokens (best) | 32,768 tokens | 131,072 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 | Yeslow · high | Yes | Yes |
| Tool calling | Yes | Yes | Yes |
| Structured output | No | Yes | No |
| Availability | |||
| Weights | Open | Open | Open |
| API model ID | step-3.5-flash-2603 | — | mimo-v2.5 |
| API providers | 3 | 2 | 21 (best) |
| Released | Apr 2, 2026 | Jun 9, 2026 | Apr 22, 2026 |
| Knowledge cutoff | Jan 2025 | — | Dec 2024 |
What would a month cost?
Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.
- Step 3.5 Flash 2603$1.60
- Gemma 4 12B IT$1.30
- MiMo-V2.5$1.96
Which should you choose?
Which is better: Step 3.5 Flash 2603, Gemma 4 12B IT or MiMo-V2.5?
MiMo-V2.5 is the better all-round choice, scoring 79/100 against Gemma 4 12B IT (75) and Step 3.5 Flash 2603 (62). It leads on inputs & features and context window. Gemma 4 12B IT wins on price. 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, Step 3.5 Flash 2603, Gemma 4 12B IT or MiMo-V2.5?
Gemma 4 12B IT is cheaper at $0.075 input / $0.275 output per million tokens (median across 2 API providers). Step 3.5 Flash 2603 costs $0.10 input / $0.30 output per million tokens (official StepFun (Global) API price); MiMo-V2.5 costs $0.14 input / $0.28 output per million tokens (official Xiaomi API price). At a typical mix of three input tokens to one output token, that is $0.125 per million tokens for Gemma 4 12B IT versus $0.15 for Step 3.5 Flash 2603 (1.2× as much) and $0.175 for MiMo-V2.5 (1.4× as much).
Which scores higher on benchmarks?
There is no independent benchmark that covers all three models yet. Step 3.5 Flash 2603 has not been scored yet, Gemma 4 12B IT has not been scored yet and MiMo-V2.5 has not been scored yet.
Which is better for coding?
There are no published SWE-bench Verified results for Step 3.5 Flash 2603, Gemma 4 12B IT and MiMo-V2.5 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?
MiMo-V2.5 has the largest context window at 1,048,576 tokens, against 262,144 for Gemma 4 12B IT and 256,000 for Step 3.5 Flash 2603. Maximum output per response: Step 3.5 Flash 2603 up to 256,000, Gemma 4 12B IT up to 32,768, MiMo-V2.5 up to 131,072 tokens.
Which can read images, PDFs, audio or video?
Step 3.5 Flash 2603 accepts text; Gemma 4 12B IT accepts text and images; MiMo-V2.5 accepts text, images, audio and video. MiMo-V2.5 handles the widest range of inputs.
Are any of these open source?
Yes, all three publish their weights, so you can self-host them.
Which is newer?
Gemma 4 12B IT is the newest, released Jun 9, 2026. MiMo-V2.5 came out Apr 22, 2026; Step 3.5 Flash 2603 came out Apr 2, 2026. Knowledge cutoff: Step 3.5 Flash 2603 Jan 2025, MiMo-V2.5 Dec 2024.
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 Oct 4, 2026.
