Sarvam 30B vs DeepSeek OCR 2 vs Granite-4.0-H-Micro
Too close to call on our weighted score (Granite-4.0-H-Micro 65, Sarvam 30B 65, DeepSeek OCR 2 32). The right pick depends on what you value most.
Sarvam AI
Sarvam 30B
65/100- ECI—
- Price$0.02 / $0.10
- Context128K
DeepSeekDeepSeek OCR 2
32/100- ECI—
- Price$0.89 / $1.47
- Context8K
IBMGranite-4.0-H-Micro
65/100- ECI—
- Price$0.017 / $0.112
- Context131K
Too close to call
It is close. Our weighted score puts them within a point (Granite-4.0-H-Micro 65/100, Sarvam 30B 65/100, DeepSeek OCR 2 32/100), so choose by what matters most for your work: Sarvam 30B on price and Granite-4.0-H-Micro for long inputs. 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 priceSarvam 30BSarvam 30B $0.04 · Granite-4.0-H-Micro $0.041 · DeepSeek OCR 2 $1.03 per 1M tokens (3:1 blend)
- Longest contextGranite-4.0-H-MicroGranite-4.0-H-Micro 131,072 · Sarvam 30B 128,000 · DeepSeek OCR 2 8,192 tokens
- Widest inputsDeepSeek OCR 2Sarvam 30B: Text · DeepSeek OCR 2: Text, Images · Granite-4.0-H-Micro: Text
- Self-hostingAll open weightsEvery model here can be downloaded and run on your own hardware
| Measure | Weight | Sarvam 30B | DeepSeek OCR 2 | Granite-4.0-H-Micro |
|---|---|---|---|---|
| Price | 50% | 100 | 49 | 100 |
| Inputs & features | 30% | 35 | 25 | 35 |
| Context window | 20% | 24 | 0 | 24 |
| Overall | 100% | 65/100 | 32/100 | 65/100 |
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 | Sarvam 30BSarvam AI | DeepSeek OCR 2DeepSeek | Granite-4.0-H-MicroIBM |
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | — | — | — |
| ECI rank | — | — | — |
| Price per million tokens | |||
| Input | $0.02 | $0.89 | $0.017 (best) |
| Output | $0.10 (best) | $1.47 | $0.112 |
| Cached input | — | — | — |
| Blended (3:1) | $0.04 (best) | $1.03 | $0.041 |
| Long-context rate | Same rate | Same rate | Same rate |
| Price source | Median of 1 providers | Median of 1 providers | Median of 1 providers |
| Limits | |||
| Context window | 128,000 tokens | 8,192 tokens | 131,072 tokens (best) |
| Max output | 128,000 tokens | 8,192 tokens | 131,072 tokens (best) |
| Inputs and features | |||
| Text | Yes | Yes | Yes |
| Images | No | Yes | No |
| PDFs | No | No | No |
| Audio | No | No | No |
| Video | No | No | No |
| Reasoning | Yes · low · medium · high | No | No |
| Tool calling | Yes | No | Yes |
| Structured output | No | No | Yes |
| Availability | |||
| Weights | Open | Open | Open |
| API model ID | sarvam-30b | — | — |
| API providers | 2 (best) | 2 (best) | 1 |
| Released | Feb 18, 2026 | Jan 27, 2026 | Oct 2, 2025 |
| 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.
- Sarvam 30B$0.40
- DeepSeek OCR 2$11.83
- Granite-4.0-H-Micro$0.394
Which should you choose?
Which is better: Sarvam 30B, DeepSeek OCR 2 or Granite-4.0-H-Micro?
It is close. Our weighted score puts them within a point (Granite-4.0-H-Micro 65/100, Sarvam 30B 65/100, DeepSeek OCR 2 32/100), so choose by what matters most for your work: Sarvam 30B on price and Granite-4.0-H-Micro for long inputs. 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, Sarvam 30B, DeepSeek OCR 2 or Granite-4.0-H-Micro?
Sarvam 30B is cheaper at $0.02 input / $0.10 output per million tokens (median across 1 API provider). Granite-4.0-H-Micro costs $0.017 input / $0.112 output per million tokens (median across 1 API provider); DeepSeek OCR 2 costs $0.89 input / $1.47 output per million tokens (median across 1 API provider). At a typical mix of three input tokens to one output token, that is $0.04 per million tokens for Sarvam 30B versus $0.041 for Granite-4.0-H-Micro (1× as much) and $1.03 for DeepSeek OCR 2 (26× as much).
Which scores higher on benchmarks?
There is no independent benchmark that covers all three models yet. Sarvam 30B has not been scored yet, DeepSeek OCR 2 has not been scored yet and Granite-4.0-H-Micro has not been scored yet.
Which is better for coding?
There are no published SWE-bench Verified results for Sarvam 30B, DeepSeek OCR 2 and Granite-4.0-H-Micro yet, so there is no like-for-like coding score. Test both on a sample of your own tasks. Note that DeepSeek OCR 2 does not support tool calling, which most coding agents need.
Which has the bigger context window?
Granite-4.0-H-Micro has the largest context window at 131,072 tokens, against 128,000 for Sarvam 30B and 8,192 for DeepSeek OCR 2. Maximum output per response: Sarvam 30B up to 128,000, DeepSeek OCR 2 up to 8,192, Granite-4.0-H-Micro up to 131,072 tokens.
Which can read images, PDFs, audio or video?
Sarvam 30B accepts text; DeepSeek OCR 2 accepts text and images; Granite-4.0-H-Micro accepts text. DeepSeek OCR 2 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?
Sarvam 30B is the newest, released Feb 18, 2026. DeepSeek OCR 2 came out Jan 27, 2026; Granite-4.0-H-Micro came out Oct 2, 2025.
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.
