Last updated:
Claude Haiku 5.5 vs Gemini 3.8 Flash
Anthropic’s Haiku 5.5 (October 7, 2026) and Google’s Gemini 3.8 Flash (September 2, 2026) are the two newest small, fast models. They are not priced in the same band: Haiku is several times cheaper per token. Here is how they compare on price, independent scores and features.
Price per 1M tokens
| Claude Haiku 5.5 | Gemini 3.8 Flash | |
|---|---|---|
| Input | $0.10 ($0.50 over 100K) | $0.75 ($1.50 from 2027) |
| Output | $0.50 ($2.50 over 100K) | $3.75 ($7.50 from 2027) |
| Cached input | $0.01 | $0.075 |
| Batch discount | 50% | 50% (Batch or Flex) |
| Free API tier | No | Yes |
Haiku 5.5 charges its higher rate for the whole request when a prompt passes 100K tokens. Gemini 3.8 Flash has one rate up to 1M tokens, but its launch price doubles on January 1, 2027.
Source: Anthropic pricing
Source: Gemini API pricing
Cost of 1,000 requests
Haiku is cheaper in every shape we priced, by the widest margin on short prompts. The gap narrows for very long prompts, where Haiku’s long-context rate applies.
| 1,000 requests of | Claude Haiku 5.5 | Gemini 3.8 Flash (2026) | Gemini 3.8 Flash (2027) |
|---|---|---|---|
| Chat reply (2K in / 500 out) | $0.45 | $3.38 | $6.75 |
| RAG answer (50K in / 1K out) | $5.50 | $41.25 | $82.50 |
| Long document (300K in / 4K out) | $160.00 | $240.00 | $480.00 |
Source: Anthropic pricing
Source: Gemini API pricing
Benchmarks: Artificial Analysis
The two vendors publish different benchmarks (Anthropic reports Humanity’s Last Exam, Google reports HLE-Verified), so their own tables cannot be lined up. Artificial Analysis runs both on the same tasks (read 2026-10-13):
| Model (effort) | Intelligence Index | Output speed | Cost per index task | Tokens to run the index |
|---|---|---|---|---|
| Claude Haiku 5.5 (Max) | 43 | 235.8 tok/s | $0.21 | 440M |
| Gemini 3.8 Flash (High) | 41 | 131.5 tok/s | $1.24 | 170M |
Haiku 5.5 scores two points higher and costs about a sixth as much per task, even though it uses more tokens. It is also faster at generating output.
Source: Artificial Analysis: Claude Haiku 5.5
Source: Artificial Analysis: Gemini 3.8 Flash
Features side by side
| Claude Haiku 5.5 | Gemini 3.8 Flash | |
|---|---|---|
| Model ID | claude-haiku-5-5 | gemini-3.8-flash |
| Context / max output | 1M / 128K | 1M / 65,536 |
| Input types | Text and images | Text, images, video, audio, PDF |
| Reasoning control | Effort: low to max | Thinking level: low, medium, high |
| Web grounding | Web search tool | Google Search and Maps grounding |
| Knowledge cutoff | June 2026 | Not published |
Source: Anthropic models overview
Source: Gemini API model page
Which one should you use?
- Pick Haiku 5.5. High-volume chat, classification, extraction and coding sub-agents on text, where price per call matters most.
- Pick Gemini 3.8 Flash. Video, audio or PDF inputs, answers grounded in Google Search, or prototyping on a free tier.
- Test both. Agent workflows: Google reports strong finance and legal agent scores for Flash; Haiku leads on the shared index. Run your own ten hardest prompts on each.
Gemini 3.8 Flash guide · Claude Haiku 5.5 guide · Sonnet 5.5 vs Gemini 3.8 Flash
FAQ
Is Claude Haiku 5.5 cheaper than Gemini 3.8 Flash?
Yes. Haiku 5.5 is $0.10 / $0.50 per 1M tokens for prompts up to 100K tokens; Gemini 3.8 Flash is $0.75 / $3.75 until the end of 2026 and $1.50 / $7.50 after.
Which is smarter, Haiku 5.5 or Gemini 3.8 Flash?
On Artificial Analysis’s Intelligence Index, Haiku 5.5 scores 43 and Gemini 3.8 Flash 41, a small gap. Results vary by task, so test on your own prompts.
Which one has a free tier?
Gemini 3.8 Flash, through the Gemini API free tier in Google AI Studio. Claude Haiku 5.5 has no free API tier.
Can Haiku 5.5 read video or audio?
No. Haiku 5.5 takes text and images. Gemini 3.8 Flash also accepts video, audio and PDF files.
Independent guide, not affiliated with Google or Anthropic. Prices and benchmarks change; check the linked official pages before you commit to a budget.