Claude Haiku 5.5
- Input / 1M tokens
- $0.10
- Output / 1M tokens
- $0.50
API models · Side by side
Claude Haiku 5.5 vs Gemini 3.8 Flash
For a text-only feature, Haiku’s lower base rates deserve attention. For a feature that reads recordings or video, Gemini’s additional input formats can change the entire workflow.
Gemini 3.8 Flash
- Input / 1M tokens
- $0.75
- Output / 1M tokens
- $3.75
The short answer
Haiku covers text, images and files. Gemini also accepts audio and video. Haiku allows a larger maximum text response, while Gemini has a slightly larger overall context window.
Compare the details
Compare context, supported inputs and usage costs. Highlighted rows show a difference.
| Feature | Claude Haiku 5.5 | Gemini 3.8 Flash |
|---|---|---|
| Input formats | file, image, text | audio, file, image, text, video |
| Output formats | text | text |
| Context / position limit | 1,000,000 | 1,048,576 |
| Maximum output tokens | 128,000 | 65,536 |
| Developer features | Reasoning output, Completion limit, Output limit, Reasoning controls, Thinking effort, Response format, Stop sequences, Structured output, Tool selection, Tool calling, Response length | Reasoning output, Output limit, Reasoning controls, Thinking effort, Response format, Seed, Stop sequences, Structured output, Temperature, Tool selection, Tool calling, Sampling controls |
| Cached input / 1M tokens | $0.01 | $0.08 |
| Cache write / 1M tokens | $0.13 | $0.04 |
| USD input / 1M tokens | $0.10 | $0.75 |
| USD output / 1M tokens | $0.50 | $3.75 |
| Model identifier | anthropic/claude-haiku-5.5 | google/gemini-3.8-flash |
API pricing
Prices are in USD per million tokens. Input is what you send; output is what the model generates. Claude and ChatGPT subscriptions are billed separately.
Claude Haiku 5.5: long-prompt rates
- Long-prompt threshold: 100,000 prompt tokens. Input: $0.50 / 1M. Output: $2.50 / 1M.
Tools, images and additional reasoning can add charges. The examples below cover input and output tokens only.
Compare local modelsWhat those prices mean in practice
Each example uses the same token budget for both models. These are calculations, not measured workloads. Cache discounts, media, tools and extra reasoning are excluded.
| Workload | Claude Haiku 5.5 | Gemini 3.8 Flash |
|---|---|---|
| 1,000 short requestsPer request: 2,000 input + 500 output tokens | $0.45 | $3.38 |
| 100 document summariesPer request: 20,000 input + 1,000 output tokens | $0.25 | $1.88 |
| 10 long-document requestsPer request: 300,000 input + 2,000 output tokens | $1.55 | $2.33 |
Estimate = requests × (input tokens × input price + output tokens × output price) ÷ 1,000,000. Long-prompt rates are applied where the pricing data provides them. Real requests can use different amounts of output.
Beyond the price tag
Start with the material you have
A meeting transcript and a meeting recording are different inputs. Haiku can work with the transcript; Gemini can accept audio directly in this comparison. A product that needs visual details from a video also needs video input or a separate frame-extraction pipeline. Decide whether those details are required before comparing token prices. Removing a conversion step may simplify the system, but it does not by itself establish a lower total bill.
Keep media costs separate from text examples
The worked costs use specified text-token counts. They are not a quote for a ten-minute recording or a video upload. Media accounting, reasoning, search and conversion services can introduce additional charges. For a transcription-and-summary product, compare the complete path: receiving the file, processing it, storing the result and returning the answer. Text-only examples are useful for the final summary step, not the whole product.
Large inputs and large outputs are different limits
A model that accepts a large document may still have a smaller maximum response. Haiku’s maximum output allowance is larger than Gemini’s here. That can matter for long reports or generated files, although shorter, separately checked sections are often easier to use. The context window must also accommodate the input and output together; do not reserve the entire allowance for uploaded material.
A sensible choice for your feature
For ticket labels, short text summaries or field extraction, start by considering the lower text budget and the response contract. For recordings, video review or mixed-media assistance, Gemini’s supported inputs provide a concrete reason to consider it. Keep accuracy checks specific to the feature: missing decisions in a meeting summary and missed visual details in a video are different failure cases.
Common questions
Does Flash mean Gemini is faster than Haiku?
The product name does not establish the latency of your requests. Request length, reasoning settings and the serving endpoint affect response time.
Can I use the text-cost examples for audio?
No. They cover the stated text-token assumptions. Audio and video workflows need their own usage accounting.