Bedrock models with vision / image input

These Amazon Bedrock models accept images as input: screenshots, scanned documents, charts, photographs and UI captures. Everything else on Bedrock is text-in, and needs an OCR step before it can read a scanned PDF.

41 of the models on Bedrock qualify today, from $0.040 per 1M input tokens. Rebuilt daily from the AWS catalogue, so this list does not go stale.

What teams get wrong about this

Images are not free and are not priced separately. Bedrock converts an image into input tokens, so a full-page screenshot can cost more than the paragraph of instructions next to it. Resize before you send: the token cost scales with resolution, and most document tasks do not need the full pixel count.

All 41 models, cheapest first

Model Input /1M Context
Gemma 3 4B IT google.gemma-3-4b-it $0.040
Nova Lite amazon.nova-lite-v1:0 $0.060 300K
Gemma 3 12B IT google.gemma-3-12b-it $0.090
Ministral 3B mistral.ministral-3-3b-instruct $0.100
Ministral 3 8B mistral.ministral-3-8b-instruct $0.150
Writer Palmyra Vision 7B writer.palmyra-vision-7b $0.150
Llama 4 Scout 17B Instruct meta.llama4-scout-17b-instruct-v1:0 $0.170
Ministral 14B 3.0 mistral.ministral-3-14b-instruct $0.200
GPT-5.6 Luna openai.gpt-5.6-luna $0.200 1000K
Gemma 3 27B PT google.gemma-3-27b-it $0.230
Llama 4 Maverick 17B Instruct meta.llama4-maverick-17b-instruct-v1:0 $0.240
Claude 3 Haiku anthropic.claude-3-haiku-20240307-v1:0 $0.250 200K
Nova 2 Lite amazon.nova-2-lite-v1:0 $0.300
Magistral Small 2509 mistral.magistral-small-2509 $0.500
Mistral Large 3 mistral.mistral-large-3-675b-instruct $0.500
Qwen3 VL 235B A22B qwen.qwen3-vl-235b-a22b $0.530
Kimi K2.5 moonshotai.kimi-k2.5 $0.600
Nova Pro amazon.nova-pro-v1:0 $0.800 300K
Claude Haiku 4.5 anthropic.claude-haiku-4-5-20251001-v1:0 $1.00 200K
Pixtral Large (25.02) mistral.pixtral-large-2502-v1:0 $2.00
GPT-5.6 Terra openai.gpt-5.6-terra $2.00 1000K
Grok 4.6 xai.grok-4.6 $2.00 500K
Nova Premier amazon.nova-premier-v1:0 $2.50 1000K
Claude 3 Sonnet anthropic.claude-3-sonnet-20240229-v1:0 $3.00 200K
Claude Sonnet 4.5 anthropic.claude-sonnet-4-5-20250929-v1:0 $3.00 200K
Claude Sonnet 4.6 anthropic.claude-sonnet-4-6 $3.00 1000K
Claude Sonnet 5 anthropic.claude-sonnet-5 $3.00 1000K
GPT-5.6 Sol openai.gpt-5.6-sol $4.00 1000K
Claude Opus 4.5 anthropic.claude-opus-4-5-20251101-v1:0 $5.00 200K
Claude Opus 4.6 anthropic.claude-opus-4-6-v1 $5.00 1000K
Claude Opus 4.7 anthropic.claude-opus-4-7 $5.00 1000K
Claude Opus 4.8 anthropic.claude-opus-4-8 $5.00 1000K
Claude Opus 5 anthropic.claude-opus-5 $5.00 1000K
Claude Fable 5 anthropic.claude-fable-5 $10.00 1000K
Claude 3.5 Sonnet anthropic.claude-3-5-sonnet-20240620-v1:0 200K
Claude 3.5 Sonnet v2 anthropic.claude-3-5-sonnet-20241022-v2:0 200K
Claude 3.7 Sonnet anthropic.claude-3-7-sonnet-20250219-v1:0 200K
Claude Opus 4.1 anthropic.claude-opus-4-1-20250805-v1:0 200K
Claude Sonnet 4 anthropic.claude-sonnet-4-20250514-v1:0 200K
NVIDIA Nemotron Nano 12B v2 VL BF16 nvidia.nemotron-nano-12b-v2
Pegasus v1.2 twelvelabs.pegasus-1-2-v1:0

Capability flags come from the AWS Bedrock model catalogue; prices from the Price List API. Region counts include cross-region inference profiles. Methodology.

Questions

Which Bedrock models can read a scanned PDF?

Any model on this page can, because a scanned PDF is a set of images. Models without vision cannot, whatever their context window — they need the document run through OCR first, which loses layout.

How are images priced on Bedrock?

As input tokens. There is no separate per-image rate on the models listed here; the image is tokenised and billed at the model input price shown in the table. That means resolution is a cost lever.