The fastest capable models in October 2026 are Gemini 3.8 Flash, Claude Haiku 5.5 and GPT-6 Luna. Artificial Analysis measured Gemini 3.8 Flash at about 302 output tokens per second, among the fastest it tracks, and it handles video and PDFs as well as text. Claude Haiku 5.5 and GPT-6 Luna are the fast picks at $0.10 per million input tokens, and both let you switch reasoning off so short requests return immediately.
The safest this comparison default, the cheaper option worth trying first, and the specialist pick — before you read the detail below.
Gemini 3.8 Flash is the strongest answer here for this comparison — pick it when quality of output matters more than the $0.75/1M/1M input you pay for it.
Claude Haiku 5.5 handles the same job for about 87% less per token. Start here and only move up if the output is not good enough.
GPT-6 Luna is the fastest of these for this comparison — worth it when latency is what the reader notices, not the last few points of reasoning depth.
Gemini 3.8 Flash is the fastest broad-use model here, at about 302 output tokens per second.
Claude Haiku 5.5 and GPT-6 Luna are the fastest cheap options, with reasoning that can be switched off.
Reasoning effort matters more to speed than the model choice — turn it down for simple requests.
Fastest is only useful if quality stays high enough for the task.
For chat interfaces and high-volume prompts, time-to-first-token matters more than peak reasoning depth.
Watch total output: Gemini 3.8 Flash is quick per token but verbose, so long answers can take longer than a terser model.
Switch the scoring lens to see whether the this comparison answer changes when cost, speed, or long-document depth leads the decision.
Google / Balanced / Oct 10, 2026
Google's fast agentic workhorse — strong coding at Flash pricing.
Ranks models by the broadest mix of coding, writing, research, and long-context usefulness.
You need long single responses (65K output cap) or your workload is output-heavy enough that its verbosity erases the per-token saving.
Where the this comparison recommendation shifts once you weigh price or latency differently.
Google's fast agentic workhorse — strong coding at Flash pricing.
Anthropic's budget model, now with adjustable reasoning, at $0.10/$0.50.
OpenAI's budget tier at half the old price.
Fastest budget multimodal model — 350 tokens/sec at Lite pricing.
SWE-bench Pro 62.5 at sixteen cents per million input.
List prices and published scores — the numbers this page's pick is built from.
| Model | Input | Output | Est. month | Context | Speed | Coding | Writing | Research |
|---|---|---|---|---|---|---|---|---|
| Gemini 3.8 FlashGoogle | $0.75/1M | $3.75/1M | $15 | 1M tokens | Very fast | 92 | 86 | 89 |
| Claude Haiku 5.5Anthropic | $0.10/1M | $0.50/1M | $2.00 | 1M tokens | Very fast | 86 | 84 | 82 |
| GPT-6 LunaOpenAI | $0.10/1M | $0.50/1M | $2.00 | 1.1M tokens | Very fast | 84 | 82 | 80 |
| Gemini 3.5 Flash-LiteGoogle | $0.30/1M | $2.50/1M | $8.00 | 1.0M tokens | Very fast | 78 | 76 | 78 |
Scores out of 100 — how we evaluate models. “Est. month” is 10M in / 2M out at list price: a ceiling, no discounts.
Why each one is on the shortlist for this comparison, what it is genuinely good at, and where we would steer you away from it.
The default answer for this comparison — 92/100 on the coding axis, and the model we would start with unless the price below rules it out.
Google's September 2, 2026 Flash model, positioned as the agentic workhorse of the Gemini 3 family — stronger coding and terminal work than 3.7 Flash at the same $0.75/$3.75 price.
You need long single responses (65K output cap) or your workload is output-heavy enough that its verbosity erases the per-token saving.
The fastest capable model in its price range. Gemini 3.8 Flash is the pick for multimodal and terminal-heavy agent work on a budget; just measure cost per task, not per token, because it writes a lot.
Full pricing, benchmark table and release notes on the Gemini 3.8 Flash page.
The value option for this comparison: about 87% less per token than Gemini 3.8 Flash, at 86/100 on coding. Worth starting here and moving up only if the output disappoints.
Anthropic's October 7, 2026 low-cost model and the first Haiku with adjustable reasoning effort — thinking can be switched off for simple requests or raised to max, at $0.10/$0.50 per 1M under 100K input tokens.
Your prompts routinely exceed 100K tokens, where the price quintuples, or the task needs sustained reasoning that Sonnet 5.5 handles far better.
The cheapest way to put an Anthropic model in a high-volume pipeline. Keep requests under 100K tokens to stay on the low rate, and set effort per task — at low effort it is a fast classifier, at max it handles real tool-use work.
Full pricing, benchmark table and release notes on the Claude Haiku 5.5 page.
The fastest model in this shortlist for this comparison. Pick it when turnaround is what your readers or users notice.
OpenAI's budget tier at half the old price. Full GPT-6 Luna review →
The alternative to check next for this comparison — 78/100 on coding.
Fastest budget multimodal model — 350 tokens/sec at Lite pricing. Full Gemini 3.5 Flash-Lite review →
UseRightAI recommendations are based on practical decision factors people actually feel in day-to-day use.
Newsletter
We email when the this comparison pick changes, when one of these models moves on price, or when something new displaces the current leader.
No spam. Useful updates only. Affiliate disclosures always clearly labeled.
Gemini 3.8 Flash, which Artificial Analysis measured at about 302 output tokens per second. Claude Haiku 5.5 and GPT-6 Luna are close behind when their reasoning is turned off.
Gemini 3.8 Flash for agentic coding — Google reports 90.8% on Terminal-Bench 2.1 — and Claude Haiku 5.5 for quick completions and small edits at a much lower price.
Often, but not always. Claude Haiku 5.5 and GPT-6 Luna are both fast and $0.10 per million input tokens. Gemini 3.8 Flash costs more at $0.75/$3.75 and is more capable.
Choose speed when the task is repetitive or low-risk. Choose quality when mistakes, rework or missed edge cases are expensive.
Two different things, and they matter in different places. Time-to-first-token is how long you wait before anything appears — it dominates how responsive a chat interface feels. Throughput is how many tokens per second arrive after that, and it dominates how long a large generation takes end to end. A model can be good at one and mediocre at the other. Our speed ratings blend both, which is why a model rated 'Very fast' is a safe pick for an interactive product and a model rated 'Deliberate' is not.
They generate hidden intermediate tokens before answering. That thinking pass is real compute you wait for and, on most providers, real tokens you pay for. It buys genuine accuracy on maths, multi-step logic, and hard debugging — and it is wasted on summarising an email. This is the single biggest speed variable in the directory: the gap between a fast non-reasoning model and a deliberate reasoning one is far larger than the gap between two models in the same class.
In batch and background work, almost entirely. If a job runs on a schedule, nobody is watching the cursor blink, so a slower and more accurate model usually wins outright. Speed earns its keep in three places: interactive chat, autocomplete-style coding assistance, and high-volume pipelines where latency multiplied across millions of calls becomes throughput you have to provision for. Outside those, optimise for correctness and cost first.