Claude Haiku 5.5 is the cheaper of these two budget models at its base rate, but Gemini 3.5 Flash-Lite has a single flat price across its 1M context. Which costs less depends on how long your prompts are.
Under 100K-token prompts, Claude Haiku 5.5 is 3× cheaper on input and 5× cheaper on output. Over 100K tokens Haiku rises to $0.50/$2.50 per 1M, at which point Gemini 3.5 Flash-Lite ($0.30/$2.50) is cheaper on input and level on output.
| Claude Haiku 5.5 | Gemini 3.5 Flash-Lite | |
|---|---|---|
| Provider | Anthropic | |
| Tier | Budget | Budget |
| Input / 1M tokens | $0.10 | $0.30 |
| Output / 1M tokens | $0.50 | $2.50 |
| Cache hit / 1M | $0.01 | $0.03 |
| Cache write / 1M | $0.125 | — |
| Context window | 1M | 1M |
| Notes | Released Oct 7, 2026; 90% cheaper than Haiku 4.5 with a 1M context window (for prompts ≤100K tokens); >100K tokens: $0.50/$2.50, cache read $0.05, cache write $0.625 | Fastest Flash tier (~350 tok/s); succeeds 3.1 Flash-Lite |
Claude Haiku 5.5 is 3.0× cheaper on input tokens ($0.10 vs $0.30 per 1M). Claude Haiku 5.5 is 5.0× cheaper on output tokens ($0.50 vs $2.50 per 1M).
| Workload | Claude Haiku 5.5 | Gemini 3.5 Flash-Lite | Difference |
|---|---|---|---|
| Support chatbot 10,000 conversations × (2,000 input + 400 output tokens) | $4.00 | $16.00 | Claude Haiku 5.5 saves $12.00 |
| Document Q&A 1,000 questions × (30,000 input + 800 output tokens) | $3.40 | $11.00 | Claude Haiku 5.5 saves $7.60 |
| Coding agent session 4M input tokens (90% cache hits) + 150,000 output tokens | $0.15 | $0.60 | Claude Haiku 5.5 saves $0.45 |
Standard rates per 1M tokens from data on this site. The coding-agent row assumes 90% of input is read from the prompt cache (full input price where a model has no cache rate). Cache-write fees, batch discounts and long-context surcharges listed in the notes are not included. Updated November 2, 2026.
most of your prompts are under 100K tokens — chat, routing, short extraction — where it is the lowest-cost option of the two by a wide margin.
you routinely send 100K–1M-token prompts and want one predictable rate, you value its listed ~350 tokens/s speed, or you want to start on Google's free tier.
Anthropic bills a Haiku 5.5 request at $0.50 input, $2.50 output and $0.05 cache read per 1M once its prompt passes 100,000 tokens, counting cached tokens. Gemini 3.5 Flash-Lite has no long-context tier on Google's pricing page. So a pipeline that summarises 300K-token filings pays less on Flash-Lite, while the cost tables above — built from short prompts — show Haiku well ahead.
Haiku's cache hit is $0.01 per 1M with writes at $0.125; Flash-Lite's cache hit is $0.03. Google also charges an hourly storage fee for explicit context caches and does not offer caching on the free tier, so heavily cached short-prompt workloads favour Haiku even more.
Google's free tier covers Flash-Lite input and output, which is useful for prototypes and low-volume tools; Anthropic's pricing page lists no free tier for Haiku 5.5. Both providers offer roughly half-price batch processing for asynchronous jobs.
For prompts up to 100K tokens, yes: $0.10/$0.50 vs $0.30/$2.50 per 1M. Above 100K tokens Flash-Lite is cheaper on input and equal on output.
Google's pricing page lists free-tier input and output for Flash-Lite; context caching is not available on the free tier.
It succeeds Gemini 3.1 Flash-Lite as Google's fastest Flash tier.
How we choose: prices come from each vendor's official pricing page and are re-checked every week. We take no affiliate fees, so rankings aren't for sale.