LightOnOCR-3 0.8B vs MiniMax-M3.1-Flash vs Nemotron 3.5 Super 120B A12B
MiniMax-M3.1-Flash comes out ahead, 66 to 45 and 30 on our weighted score, and it is the cheaper option too.
LightOn
LightOnOCR-3 0.8B
30/100- ECI—
- Price—
- Context262K
- Our pick
MiniMaxMiniMax-M3.1-Flash
66/100- ECI—
- Price$0.15 / $0.60
- Context1M
NVIDIANemotron 3.5 Super 120B A12B
45/100- ECI—
- Price$0.20 / $0.50
- Context1.05M
MiniMax-M3.1-Flash is our pick
MiniMax-M3.1-Flash is the better all-round choice, scoring 66/100 against Nemotron 3.5 Super 120B A12B (45) and LightOnOCR-3 0.8B (30). It leads on inputs & features. The score weighs inputs & features 60%, context window 40%. 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 priceMiniMax-M3.1-FlashMiniMax-M3.1-Flash $0.263 · Nemotron 3.5 Super 120B A12B $0.275 per 1M tokens (3:1 blend) · LightOnOCR-3 0.8B unpriced
- Longest contextNemotron 3.5 Super 120B A12BNemotron 3.5 Super 120B A12B 1,048,576 · MiniMax-M3.1-Flash 1,000,000 · LightOnOCR-3 0.8B 262,144 tokens
- Widest inputsMiniMax-M3.1-FlashLightOnOCR-3 0.8B: Text, Images · MiniMax-M3.1-Flash: Text, Images, Video · Nemotron 3.5 Super 120B A12B: Text
- Self-hostingLightOnOCR-3 0.8B and Nemotron 3.5 Super 120B A12BPublishes downloadable weights (apache-2.0)
| Measure | Weight | LightOnOCR-3 0.8B | MiniMax-M3.1-Flash | Nemotron 3.5 Super 120B A12B |
|---|---|---|---|---|
| Inputs & features | 60% | 25 | 70 | 35 |
| Context window | 40% | 37 | 60 | 61 |
| Overall | 100% | 30/100 | 66/100 | 45/100 |
Left out because at least one model lacks the data: capability and price. The remaining weights were rescaled.
Every spec in one table
Highlighted cells lead their row. Dashes mean the data is not published.
| Specification | LightOnOCR-3 0.8BLightOn | MiniMax-M3.1-FlashMiniMax | Nemotron 3.5 Super 120B A12BNVIDIA |
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | — | — | — |
| ECI rank | — | — | — |
| Price per million tokens | |||
| Input | — | $0.15 (best) | $0.20 |
| Output | — | $0.60 | $0.50 (best) |
| Cached input | — | $0.015 | — |
| Blended (3:1) | — | $0.263 (best) | $0.275 |
| Long-context rate | — | Same rate | Same rate |
| Price source | — | Official MiniMax (minimax.io) API | Median of 1 providers |
| Limits | |||
| Context window | 262,144 tokens | 1,000,000 tokens | 1,048,576 tokens (best) |
| Max output | 16,384 tokens | 512,000 tokens (best) | 131,072 tokens |
| Inputs and features | |||
| Text | Yes | Yes | Yes |
| Images | Yes | Yes | No |
| PDFs | No | No | No |
| Audio | No | No | No |
| Video | No | Yes | No |
| Reasoning | No | Yeslow · medium · high · xhigh · max | Yes |
| Tool calling | No | Yes | Yes |
| Structured output | No | No | No |
| Availability | |||
| Weights | Openapache-2.0 | Proprietary | Open |
| API model ID | — | MiniMax-M3.1-Flash | — |
| API providers | 1 | 25 (best) | 1 |
| Released | Oct 8, 2026 | Oct 19, 2026 | Oct 15, 2026 |
| Knowledge cutoff | — | — | — |
What would a month cost?
Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.
- LightOnOCR-3 0.8B—
- MiniMax-M3.1-Flash$2.70
- Nemotron 3.5 Super 120B A12B$3.00
Which should you choose?
Which is better: LightOnOCR-3 0.8B, MiniMax-M3.1-Flash or Nemotron 3.5 Super 120B A12B?
MiniMax-M3.1-Flash is the better all-round choice, scoring 66/100 against Nemotron 3.5 Super 120B A12B (45) and LightOnOCR-3 0.8B (30). It leads on inputs & features. The score weighs inputs & features 60%, context window 40%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.
Which is cheaper, LightOnOCR-3 0.8B, MiniMax-M3.1-Flash or Nemotron 3.5 Super 120B A12B?
MiniMax-M3.1-Flash is cheaper at $0.15 input / $0.60 output per million tokens (official MiniMax (minimax.io) API price). Nemotron 3.5 Super 120B A12B costs $0.20 input / $0.50 output per million tokens (median across 1 API provider). At a typical mix of three input tokens to one output token, that is $0.263 per million tokens for MiniMax-M3.1-Flash versus $0.275 for Nemotron 3.5 Super 120B A12B (1× as much). LightOnOCR-3 0.8B has no published per-token price.
Which scores higher on benchmarks?
There is no independent benchmark that covers all three models yet. LightOnOCR-3 0.8B has not been scored yet, MiniMax-M3.1-Flash has not been scored yet and Nemotron 3.5 Super 120B A12B has not been scored yet.
Which is better for coding?
There are no published SWE-bench Verified results for LightOnOCR-3 0.8B, MiniMax-M3.1-Flash and Nemotron 3.5 Super 120B A12B yet, so there is no like-for-like coding score. Test both on a sample of your own tasks. Note that LightOnOCR-3 0.8B does not support tool calling, which most coding agents need.
Which has the bigger context window?
Nemotron 3.5 Super 120B A12B has the largest context window at 1,048,576 tokens, against 1,000,000 for MiniMax-M3.1-Flash and 262,144 for LightOnOCR-3 0.8B. Maximum output per response: LightOnOCR-3 0.8B up to 16,384, MiniMax-M3.1-Flash up to 512,000, Nemotron 3.5 Super 120B A12B up to 131,072 tokens.
Which can read images, PDFs, audio or video?
LightOnOCR-3 0.8B accepts text and images; MiniMax-M3.1-Flash accepts text, images and video; Nemotron 3.5 Super 120B A12B accepts text. MiniMax-M3.1-Flash handles the widest range of inputs.
Are any of these open source?
LightOnOCR-3 0.8B and Nemotron 3.5 Super 120B A12B publishes its weights (apache-2.0) and can be self-hosted; MiniMax-M3.1-Flash is proprietary.
Which is newer?
MiniMax-M3.1-Flash is the newest, released Oct 19, 2026. Nemotron 3.5 Super 120B A12B came out Oct 15, 2026; LightOnOCR-3 0.8B came out Oct 8, 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.
