ZMIME
Comparison · 3 models · Updated Nov 18, 2026

Pixtral 12B vs Nova Lite vs Llama-3.2-11B-Vision-Instruct

No independent capability benchmark covers every model here yet. The score uses price, inputs & features and context window.

Price in rating: Included · Exclude price

  1. Mistral AI

    Pixtral 12B

    Released Sep 11, 2024Deprecated

    64/100
    • ECI—
    • Price$0.15 / $0.15
    • Context128K
  2. Amazon

    Nova Lite

    Released Dec 3, 2024

    76/100
    • ECI—
    • Price$0.06 / $0.24
    • Context300K
  3. Meta

    Llama-3.2-11B-Vision-Instruct

    Released Sep 25, 2024

    66/100
    • ECI—
    • Price$0.055 / $0.345
    • Context128K
01 — Verdict

Capability evidence is incomplete

There is not enough shared capability evidence to name an overall winner. The scores (Nova Lite 76/100, Llama-3.2-11B-Vision-Instruct 66/100, Pixtral 12B 64/100) compare the available specifications and 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 priceNova LiteNova Lite $0.105 · Llama-3.2-11B-Vision-Instruct $0.128 · Pixtral 12B $0.15 per 1M tokens (3:1 blend)
  • Longest contextNova LiteNova Lite 300,000 · Pixtral 12B 128,000 · Llama-3.2-11B-Vision-Instruct 128,000 tokens
  • Widest inputsNova LitePixtral 12B: Text, Images · Nova Lite: Text, Images, PDFs, Video · Llama-3.2-11B-Vision-Instruct: Text, Images
  • Self-hostingPixtral 12B and Llama-3.2-11B-Vision-InstructPublishes downloadable weights
How the score is built
MeasureWeightPixtral 12BNova LiteLlama-3.2-11B-Vision-Instruct
Price50%859288
Inputs & features30%507050
Context window20%324532
Specs & price only100%64/10076/10066/100

Left out because at least one model lacks the data: capability. The remaining weights were rescaled.

02 — Side by side

Every spec in one table

Highlighted cells lead their row. Dashes mean the data is not published.

Pixtral 12B vs Nova Lite vs Llama-3.2-11B-Vision-Instruct specifications side by side
SpecificationPixtral 12BMistral AINova LiteAmazonLlama-3.2-11B-Vision-InstructMeta
Capability
Capabilities Index (ECI)———
ECI rank———
Price per million tokens
Input$0.15$0.06$0.055 (best)
Output$0.15 (best)$0.24$0.345
Cached input—$0.015—
Blended (3:1)$0.15$0.105 (best)$0.128
Long-context rateSame rateSame rateSame rate
Price sourceOfficial Mistral APIOfficial Amazon Bedrock APIMedian of 3 providers
Limits
Context window128,000 tokens300,000 tokens (best)128,000 tokens
Max output128,000 tokens (best)10,000 tokens4,096 tokens
Inputs and features
TextYesYesYes
ImagesYesYesYes
PDFsNoYesNo
AudioNoNoNo
VideoNoYesNo
ReasoningNoNoNo
Tool callingYesYesYes
Structured outputNoNoNo
Availability
WeightsOpenProprietaryOpen
API model IDpixtral-12bamazon.nova-lite-v1:0—
API providers28 (best)4
ReleasedSep 11, 2024Dec 3, 2024Sep 25, 2024
Knowledge cutoffSep 2024Oct 2024Dec 2023
03 — Cost

What would a month cost?

Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.

  • Pixtral 12B$1.80
  • Nova Lite$1.08
  • Llama-3.2-11B-Vision-Instruct$1.24
04 — Questions

Which should you choose?

Which is better: Pixtral 12B, Nova Lite or Llama-3.2-11B-Vision-Instruct?

There is not enough shared capability evidence to name an overall winner. The scores (Nova Lite 76/100, Llama-3.2-11B-Vision-Instruct 66/100, Pixtral 12B 64/100) compare the available specifications and 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, Pixtral 12B, Nova Lite or Llama-3.2-11B-Vision-Instruct?

Nova Lite is cheaper at $0.06 input / $0.24 output per million tokens (official Amazon Bedrock API price). Llama-3.2-11B-Vision-Instruct costs $0.055 input / $0.345 output per million tokens (median across 3 API providers); Pixtral 12B costs $0.15 input / $0.15 output per million tokens (official Mistral API price). At a typical mix of three input tokens to one output token, that is $0.105 per million tokens for Nova Lite versus $0.128 for Llama-3.2-11B-Vision-Instruct (1.2× as much) and $0.15 for Pixtral 12B (1.4× as much).

Which scores higher on benchmarks?

There is no independent benchmark that covers all three models yet. Pixtral 12B has not been scored yet, Nova Lite has not been scored yet and Llama-3.2-11B-Vision-Instruct has not been scored yet.

Which is better for coding?

There are no published SWE-bench Verified results for Pixtral 12B, Nova Lite and Llama-3.2-11B-Vision-Instruct 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?

Nova Lite has the largest context window at 300,000 tokens, against 128,000 for Pixtral 12B and 128,000 for Llama-3.2-11B-Vision-Instruct. Maximum output per response: Pixtral 12B up to 128,000, Nova Lite up to 10,000, Llama-3.2-11B-Vision-Instruct up to 4,096 tokens.

Which can read images, PDFs, audio or video?

Pixtral 12B accepts text and images; Nova Lite accepts text, images, PDFs and video; Llama-3.2-11B-Vision-Instruct accepts text and images. Nova Lite handles the widest range of inputs.

Are any of these open source?

Pixtral 12B and Llama-3.2-11B-Vision-Instruct publishes its weights and can be self-hosted; Nova Lite is proprietary.

Which is newer?

Nova Lite is the newest, released Dec 3, 2024. Llama-3.2-11B-Vision-Instruct came out Sep 25, 2024; Pixtral 12B came out Sep 11, 2024. Knowledge cutoff: Pixtral 12B Sep 2024, Nova Lite Oct 2024, Llama-3.2-11B-Vision-Instruct Dec 2023.

How do you decide the winner?

This rating uses price (50%, blended price per million tokens (3 input : 1 output), log scale); inputs & features (30%, image, PDF, audio and video input, tool calling, structured output and reasoning); context window (20%, 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 18, 2026.