Llama-3.2-90B-Vision-Instruct vs Gemma 3 4B IT vs Claude Haiku 3.5
Claude Haiku 3.5 comes out ahead, 48 to 44 and 36 on our weighted score, though Gemma 3 4B IT is 26× cheaper per token.
MetaLlama-3.2-90B-Vision-Instruct
44/100- ECI125.5
- Price—
- Context128K
GoogleGemma 3 4B IT
36/100- ECI115.8
- Price$0.05 / $0.10
- Context131K
- Our pick
AnthropicClaude Haiku 3.5
48/100- ECI127.0
- Price$0.80 / $4.00
- Context200K
Claude Haiku 3.5 is our pick
Claude Haiku 3.5 is the better all-round choice, scoring 48/100 against Llama-3.2-90B-Vision-Instruct (44) and Gemma 3 4B IT (36). It leads on capability, inputs & features and context window. The score weighs capability 67%, inputs & features 20%, context window 13%.
- CapabilityClaude Haiku 3.5Capabilities Index (ECI): Claude Haiku 3.5 127.0 · Llama-3.2-90B-Vision-Instruct 125.5 · Gemma 3 4B IT 115.8
- Lowest priceGemma 3 4B ITGemma 3 4B IT $0.063 · Claude Haiku 3.5 $1.60 per 1M tokens (3:1 blend) · Llama-3.2-90B-Vision-Instruct unpriced
- Longest contextClaude Haiku 3.5Claude Haiku 3.5 200,000 · Gemma 3 4B IT 131,072 · Llama-3.2-90B-Vision-Instruct 128,000 tokens
- Widest inputsClaude Haiku 3.5Llama-3.2-90B-Vision-Instruct: Text, Images · Gemma 3 4B IT: Text, Images · Claude Haiku 3.5: Text, Images, PDFs
- Self-hostingLlama-3.2-90B-Vision-Instruct and Gemma 3 4B ITPublishes downloadable weights (Llama 3.2 Community License)
| Measure | Weight | Llama-3.2-90B-Vision-Instruct | Gemma 3 4B IT | Claude Haiku 3.5 |
|---|---|---|---|---|
| CapabilityCapabilities Index (ECI) | 67% | 46 | 34 | 48 |
| Inputs & features | 20% | 50 | 50 | 60 |
| Context window | 13% | 24 | 25 | 32 |
| Overall | 100% | 44/100 | 36/100 | 48/100 |
Left out because at least one model lacks the data: price. The remaining weights were rescaled.
Every spec in one table
Highlighted cells lead their row. Dashes mean the data is not published.
| Specification | Llama-3.2-90B-Vision-InstructMeta | Gemma 3 4B ITGoogle | Claude Haiku 3.5Anthropic |
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | 125.5 | 115.8 | 127.0 (best) |
| ECI rank | #150 of 165 | #163 of 165 | #147 of 165 (best) |
| GPQA DiamondGraduate-level science questions | 41.0% (best) | 23.2% | 38.1% |
| OTIS Mock AIME 2024–2025Competition mathematics | 2.6% | 7.5% (best) | 4.3% |
| Price per million tokens | |||
| Input | — | $0.05 (best) | $0.80 |
| Output | — | $0.10 (best) | $4.00 |
| Cached input | — | — | — |
| Blended (3:1) | — | $0.063 (best) | $1.60 |
| Long-context rate | — | Same rate | Same rate |
| Price source | — | Median of 9 providers | Median of 3 providers |
| Limits | |||
| Context window | 128,000 tokens | 131,072 tokens | 200,000 tokens (best) |
| Max output | 4,096 tokens | 131,072 tokens (best) | 8,192 tokens |
| Inputs and features | |||
| Text | Yes | Yes | Yes |
| Images | Yes | Yes | Yes |
| PDFs | No | No | Yes |
| Audio | No | No | No |
| Video | No | No | No |
| Reasoning | No | No | No |
| Tool calling | Yes | Yes | Yes |
| Structured output | No | No | No |
| Availability | |||
| Weights | OpenLlama 3.2 Community License | Open | Proprietary |
| API model ID | — | — | — |
| API providers | 1 | 10 (best) | 4 |
| Released | Sep 25, 2024 | Mar 12, 2025 | Oct 22, 2024 |
| Knowledge cutoff | Dec 2023 | Aug 2024 | Jul 31, 2024 |
What would a month cost?
Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.
- Llama-3.2-90B-Vision-Instruct—
- Gemma 3 4B IT$0.70
- Claude Haiku 3.5$16.00
Which should you choose?
Which is better: Llama-3.2-90B-Vision-Instruct, Gemma 3 4B IT or Claude Haiku 3.5?
Claude Haiku 3.5 is the better all-round choice, scoring 48/100 against Llama-3.2-90B-Vision-Instruct (44) and Gemma 3 4B IT (36). It leads on capability, inputs & features and context window. The score weighs capability 67%, inputs & features 20%, context window 13%.
Which is cheaper, Llama-3.2-90B-Vision-Instruct, Gemma 3 4B IT or Claude Haiku 3.5?
Gemma 3 4B IT is cheaper at $0.05 input / $0.10 output per million tokens (median across 9 API providers). Claude Haiku 3.5 costs $0.80 input / $4.00 output per million tokens (median across 3 API providers). At a typical mix of three input tokens to one output token, that is $0.063 per million tokens for Gemma 3 4B IT versus $1.60 for Claude Haiku 3.5 (26× as much). Llama-3.2-90B-Vision-Instruct has no published per-token price.
Which scores higher on benchmarks?
Claude Haiku 3.5 scores higher on the Capabilities Index (ECI): Claude Haiku 3.5 127.0 (#147 of 165), Llama-3.2-90B-Vision-Instruct 125.5 (#150 of 165) and Gemma 3 4B IT 115.8 (#163 of 165). The confidence ranges of the top two overlap (121.4–129.2 vs 120.3–127.4), so treat the gap as small. On individual benchmarks: GPQA Diamond — Llama-3.2-90B-Vision-Instruct 41.0%, Claude Haiku 3.5 38.1%, Gemma 3 4B IT 23.2%; OTIS Mock AIME 2024–2025 — Gemma 3 4B IT 7.5%, Claude Haiku 3.5 4.3%, Llama-3.2-90B-Vision-Instruct 2.6%.
Which is better for coding?
There are no published SWE-bench Verified results for Llama-3.2-90B-Vision-Instruct, Gemma 3 4B IT and Claude Haiku 3.5 yet, so there is no like-for-like coding score. On overall capability, Claude Haiku 3.5 leads, which tends to carry over to coding, but test on your own codebase. All three support tool calling for agent workflows.
Which has the bigger context window?
Claude Haiku 3.5 has the largest context window at 200,000 tokens, against 131,072 for Gemma 3 4B IT and 128,000 for Llama-3.2-90B-Vision-Instruct. Maximum output per response: Llama-3.2-90B-Vision-Instruct up to 4,096, Gemma 3 4B IT up to 131,072, Claude Haiku 3.5 up to 8,192 tokens.
Which can read images, PDFs, audio or video?
Llama-3.2-90B-Vision-Instruct accepts text and images; Gemma 3 4B IT accepts text and images; Claude Haiku 3.5 accepts text, images and PDFs. Claude Haiku 3.5 handles the widest range of inputs.
Are any of these open source?
Llama-3.2-90B-Vision-Instruct and Gemma 3 4B IT publishes its weights (Llama 3.2 Community License) and can be self-hosted; Claude Haiku 3.5 is proprietary.
Which is newer?
Gemma 3 4B IT is the newest, released Mar 12, 2025. Claude Haiku 3.5 came out Oct 22, 2024; Llama-3.2-90B-Vision-Instruct came out Sep 25, 2024. Knowledge cutoff: Llama-3.2-90B-Vision-Instruct Dec 2023, Gemma 3 4B IT Aug 2024, Claude Haiku 3.5 Jul 31, 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 Nov 2, 2026.
Related comparisons
- Llama-3.2-90B-Vision-Instruct vs GPT-4o miniOpenAI · ECI 126.5
- Llama-3.2-90B-Vision-Instruct vs Mistral Large 2407Mistral AI · ECI 127.6
- Llama-3.2-90B-Vision-Instruct vs Gemma 2 27BGoogle · ECI 122.1
- Gemma 3 4B IT vs Llama-3.2-1BMeta
- Gemma 3 4B IT vs Qwen2.5 7B InstructAlibaba (Qwen) · ECI 118.5
- Gemma 3 4B IT vs Mistral NemoMistral AI · ECI 118.6
