Answers at the
speed of thought.typing.light.instinct.
The fastest model we ship — sharper than what you pay $20 for. Free to start. Pro is five dollars.
Clove Code
187msTTFT
72%SWE-bench
$5Pro / mo
187ms
ttft
TTFT — 1.4× faster than Claude Haiku 4.5, the nearest fast-chat competitor
+2 pts
swe-bench
Clove Sky leads Claude Sonnet 4.6 on fast-tier SWE-bench Verified, at 1.4× the TTFT
$5
pro
Pro is five dollars a month — not twenty. Coding agents included, no surprise bills.
Artificial Analysis Speed Index
The numbers
are not close.
We benchmark Clove Dojo against the fast chatbots people actually reach for day to day — same prompts, same hardware, every run. Dojo leads on TTFT and latency; Haiku pushes it hardest on throughput.
- Clove Dojo89
- Claude Haiku 4.581
- Gemini 2.5 Flash63
- ChatGPT51
Time to first token
TTFTClove Dojo187ms
Claude Haiku 4.5268ms
Gemini 2.5 Flash341ms
ChatGPT397ms
Full reply (512 tokens)
E2E latency (512 tok)Clove Dojo4.1sec
Claude Haiku 4.55.6sec
Gemini 2.5 Flash7.2sec
ChatGPT8.9sec
Sustained decode throughput
Output throughputClove Dojo217tok/s
Claude Haiku 4.5203tok/s
Gemini 2.5 Flash148tok/s
ChatGPT115tok/s
1.4×
faster first token than Claude Haiku 4.5, the nearest fast-chat competitor — with Haiku closing the gap to 7% on sustained throughput.
SWE-bench Verified
Speed doesn't
cost intelligence.
Architect is measured against the top frontier models. Sky is measured against the leading fast-tier models — the ones you actually reach for when you need speed without sacrificing accuracy. Same prompts, graders, and hardware every run.
- Claude Opus 4.876
- GPT-5.574
- Clove Architect73
- Claude Sonnet 4.671
- Gemini 3.1 Pro68
Higher is better — SWE-bench Verified v1.2, n=500 tasks, ±1.4 pts at 95% CI
- Clove Sky82
- Claude Sonnet 4.679
- Gemini 3.1 Pro73
Clove Sky
Fast coding agent
Debug & explain
SWE-bench LiteClove Sky52%
Claude Sonnet 4.654%
Gemini 3.1 Pro49%
SWE-bench Lite — 300 real-repo issues, blind pass/fail graders
Ship a feature
SWE-bench VerifiedClove Sky72%
Claude Sonnet 4.670%
Gemini 3.1 Pro67%
SWE-bench Verified — spec to green CI, median of 3 attempts
Speed-adjusted coding
Aider PolyglotClove Sky82idx
Claude Sonnet 4.678idx
Gemini 3.1 Pro68idx
Aider Polyglot — SWE-bench score × normalized TTFT (not raw IQ)
Clove Architect
Deep coding agent
Multi-file refactor
CanItEditClove Architect80%
GPT-5.582%
Claude Opus 4.878%
Claude Sonnet 4.674%
Gemini 3.1 Pro70%
CanItEdit — 80 multi-file edits, regression suite must stay green
Architecture & tradeoffs
GPQA DiamondClove Architect77%
GPT-5.581%
Claude Opus 4.878%
Claude Sonnet 4.673%
Gemini 3.1 Pro69%
GPQA Diamond — rubric scored 1–5 by reviewers blind to model
Hard problem pass rate
LiveCodeBenchClove Architect75%
GPT-5.580%
Claude Opus 4.877%
Claude Sonnet 4.672%
Gemini 3.1 Pro67%
LiveCodeBench — 120 production-grade tasks, pass@1
+2
Two points ahead of the nearest fast-tier competitor on SWE-bench Verified — while running 1.4× faster on TTFT in the same published runs.
Why Clove
Fast, sharp, and priced right.
First token in 187ms
Our ledger puts Clove Dojo at 1.4× faster than Claude Haiku 4.5 on time-to-first-token, with Haiku narrowing the gap on throughput.
Intelligence that keeps up
Clove Sky leads Claude Sonnet 4.6 on fast-tier SWE-bench Verified — at a fraction of the latency.
Two coding agents, one price
Clove Sky and Clove Architect are both included in Pro — no separate subscriptions, no per-agent add-ons.
Usage ledger
More runway
per dollar.
We publish every tier transparently. Pro gives you more headroom across chat and coding than typical Plus tiers that hide behind opaque or windowed limits.
- Clove Pro
- Clove Forge
- ChatGPT Plus
- Claude Pro
Relative chat runway per dollar — longer bar means more headroom
Chat messages per day
Clove Pro
Clove Forge
ChatGPT Plus
Claude Pro
Clove Sky agent runs / week
Clove Pro
Clove Forge
Claude Code
Cursor Pro
Combined coding runs / week
Clove Pro
Clove Forge
Claude Code
Cursor Pro
Chat messages per dollar / mo
Clove Pro
Clove Forge
ChatGPT Plus
Claude Pro
Forge and Summit add significantly more headroom than Pro — without stacking Claude Code and ChatGPT Plus at higher monthly cost with lower, opaque limits.
Clove tiers are fixed and published. Competitor bars reflect publicly documented tiers or commonly reported medians. Your mileage varies; we show relative runway, not marketing adjectives.
Clove Code
Two coding agents.
Included in Pro.
Purpose-built for developers — full access to Sky and Architect, published transparently in the usage ledger. Claude Code charges extra for a smaller, opaque pool. Clove Code is included in Pro.
Clove Sky
Fast coding
Debug, snippets, explain, ship.
Enough for most daily coding.
Clove Architect
Deep coding
Architecture, refactors, tradeoffs.
Reach for this when the problem is hard.
Need more headroom?
Pro covers most daily development. Forge and Summit add expanded limits for all-day coding and the heaviest agentic workflows.
Compare Forge & Summit →macOS (.dmg) and Windows. A paid plan (Pro, Forge, or Summit) unlocks coding agents.
Pricing
Start for $2.50. Then $5/mo.
First month at half price. No tricks, no surprise bills. Most developers should stop at Pro.
Free
$0 / forever
A real free plan you can use every day — not a three-message demo that locks you out.
- Daily use at no cost
- Full access to Clove Dojo
- Private conversations
- No credit card
Pro
$5 / month
Clove Code included — enough for most daily development. $2.50 first month, then $5/mo.
- $2.50 first month — then $5/mo
- Full Clove chat access
- Clove Sky — fast coding agent
- Clove Architect — deep coding agent
- Generous coding headroom
- Saved chat history
- Priority responses
- Longer conversations
Forge
$15 / month
Expanded limits for long coding sessions and heavier agentic workflows when Pro isn't quite enough.
- Everything in Pro
- Expanded chat headroom
- Higher Sky capacity
- Higher Architect capacity
- Built for all-day coding
Summit
$40 / month
Our highest tier — maximum published limits for power users and teams who live in Clove Code.
- Everything in Forge
- Maximum published limits
- Highest Sky capacity
- Highest Architect capacity
- For teams and power users