Every monthly plan that includes Kimi K3, converted to requests per $1 at the model's typical request shape. Same engine as the main calculator — proxy estimates carry a ~ marker; coarse estimates carry ≈.
Kimi K3 is Moonshot's flagship model for agentic coding and reasoning.
Claude Opus 5GPT-5.6 SolGemini 3.7 FlashClaude Sonnet 5Claude Fable 5.1Claude Fable 5Grok 4.6Kimi K3GPT-5.6 TerraGPT-5.6 LunaGPT-5.5GPT-5.4Claude Opus 4.8Claude Opus 4.7Claude Haiku 4.5Claude Sonnet 4.6Kimi K2.7 CodeGemini 3.6 FlashGemini 3.5 FlashGrok 4.5GPT-5 MiniGPT-5.4 Mini
Ceiling
$200/mo total AI credits · 2.9× Pro+ usage
Reachability
—
Pricing notes
—
Cost rule
~proxy estimate · platform.moonshot.cn/docs/pricing/chat — Kimi K3 row ($3/$15/$0.30; vendor list rate, matches the opencode-go / cline-pass rows in this repo); Copilot prices premium usage at provider list rates
Claude Opus 5GPT-5.6 SolGemini 3.7 FlashClaude Sonnet 5Claude Fable 5.1Claude Fable 5Grok 4.6Kimi K3GPT-5.6 TerraGPT-5.6 LunaGPT-5.5GPT-5.4Claude Opus 4.8Claude Opus 4.7Claude Haiku 4.5Claude Sonnet 4.6Kimi K2.7 CodeGemini 3.6 FlashGemini 3.5 FlashGrok 4.5GPT-5 MiniGPT-5.4 Mini
Ceiling
$70/mo total AI credits · 4× Pro usage
Reachability
—
Pricing notes
—
Cost rule
~proxy estimate · platform.moonshot.cn/docs/pricing/chat — Kimi K3 row ($3/$15/$0.30; vendor list rate, matches the opencode-go / cline-pass rows in this repo); Copilot prices premium usage at provider list rates
~proxy estimate · platform.moonshot.cn/docs/pricing/chat — Kimi K3 row ($3/$15/$0.30; vendor list rate, matches the opencode-go / cline-pass rows in this repo); Copilot prices premium usage at provider list rates
$269/mo compute budget · no 5h/week windows · highest-priority scheduling
Reachability
Any agent · OpenAI-compatible key
Pricing notes
Dollar budget spent at provider list rates — provider claims it does the work of up to $880/mo of direct API spend; no overage charges.
Cost rule
~proxy estimate · platform.moonshot.ai/docs/pricing/chat (.md) — Kimi K3 row ($3.00 input / $15.00 output / $0.30 cache-hit input; the CNY mirror platform.kimi.com prints the same row at ¥20/¥100/¥2; read 2026-09-24)
Standard ComputeStandard$95/mo compute budget · no 5h/week windows · priority scheduling
$95/mo compute budget · no 5h/week windows · priority scheduling
Reachability
Any agent · OpenAI-compatible key
Pricing notes
Dollar budget spent at provider list rates — provider claims it does the work of up to $310/mo of direct API spend; no overage charges.
Cost rule
~proxy estimate · platform.moonshot.ai/docs/pricing/chat (.md) — Kimi K3 row ($3.00 input / $15.00 output / $0.30 cache-hit input; the CNY mirror platform.kimi.com prints the same row at ¥20/¥100/¥2; read 2026-09-24)
Standard ComputeStarter$20/mo compute budget · no 5h/week windows · one agent at a time
$20/mo compute budget · no 5h/week windows · one agent at a time
Reachability
Any agent · OpenAI-compatible key
Pricing notes
Dollar budget spent at provider list rates — provider claims it does the work of up to $60/mo of direct API spend; no overage charges.
Cost rule
~proxy estimate · platform.moonshot.ai/docs/pricing/chat (.md) — Kimi K3 row ($3.00 input / $15.00 output / $0.30 cache-hit input; the CNY mirror platform.kimi.com prints the same row at ¥20/¥100/¥2; read 2026-09-24)
Standard ComputeEconomy$41/mo compute budget · no 5h/week windows · parallel agents
$41/mo compute budget · no 5h/week windows · parallel agents
Reachability
Any agent · OpenAI-compatible key
Pricing notes
Dollar budget spent at provider list rates — provider claims it does the work of up to $130/mo of direct API spend; no overage charges.
Cost rule
~proxy estimate · platform.moonshot.ai/docs/pricing/chat (.md) — Kimi K3 row ($3.00 input / $15.00 output / $0.30 cache-hit input; the CNY mirror platform.kimi.com prints the same row at ¥20/¥100/¥2; read 2026-09-24)
All models (Max and Ultra currently cover nearly identical lists — the difference is quota, not model selection). Non-production use (personal dev / prototyping — PAYG is the production tier). Flow ≈ $0.03283 (floats live); per-model Flow burn and per-tier model lists are not published in text. Provider "Worth" claim: $480/mo equivalent API value (≈2.4× the fee), from its published equivalent-pay-as-you-go table.
Cost rule
≈coarse estimate · zenmux.ai/api/v1/models — MoonshotAI: Kimi K3 ($3.00/$15.00/$0.30 cached read; the row previously stored 0.9x these values; read 2026-09-24)
Basic + premium models, covering the mainstream flagships. Non-production use (personal dev / prototyping — PAYG is the production tier). Flow ≈ $0.03283 (floats live); per-model Flow burn and per-tier model lists are not published in text. Provider "Worth" claim: $180/mo equivalent API value (≈1.8× the fee), from its published equivalent-pay-as-you-go table.
Cost rule
≈coarse estimate · zenmux.ai/api/v1/models — MoonshotAI: Kimi K3 ($3.00/$15.00/$0.30 cached read; the row previously stored 0.9x these values; read 2026-09-24)
CommandCodeProviderPAYG · top-ups roll over, never expire · zero markup
Basic models plus rotating limited-time premium models. Non-production use (personal dev / prototyping — PAYG is the production tier). Flow ≈ $0.03283 (floats live); per-model Flow burn and per-tier model lists are not published in text. Provider "Worth" claim: $30/mo equivalent API value (≈1.5× the fee), from its published equivalent-pay-as-you-go table.
Cost rule
≈coarse estimate · zenmux.ai/api/v1/models — MoonshotAI: Kimi K3 ($3.00/$15.00/$0.30 cached read; the row previously stored 0.9x these values; read 2026-09-24)
$70 in credits/mo · + Smart Compression, Analytics Pro, Privacy Pack, 10 extra API keys, Custom Router, Fusion
Reachability
Any agent · OpenAI-compatible key
Pricing notes
Includes Custom Router, so you can pin a lineup and Ozore routes only among your picks. Cached-read pricing is not published, so the calculator estimates it at 10% of the input rate.
Cost rule
~proxy estimate · floor: cached reads bill at the input rate · ozore.com/api/credits/pricing (live 2026-09-22, billed row moonshotai/kimi-k3 $2.10/10.50 per 1M = 30% off list $3.00/15.00); cached-read not published → cached tokens billed at the input rate
OzoreBasic$20 in credits/mo · spend on any model at discounted rates · cancel anytime
$20 in credits/mo · spend on any model at discounted rates · cancel anytime
Reachability
Any agent · OpenAI-compatible key
Pricing notes
Requests are auto-routed — you do not pick the model, so a per-model figure is an upper bound on this tier; Direct Pin and Custom Router are Pro features. Cached-read pricing is not published, so the calculator estimates it at 10% of the input rate.
Cost rule
~proxy estimate · floor: cached reads bill at the input rate · ozore.com/api/credits/pricing (live 2026-09-22, billed row moonshotai/kimi-k3 $2.10/10.50 per 1M = 30% off list $3.00/15.00); cached-read not published → cached tokens billed at the input rate
Bills $3.99/mo after the free month · cancel anytime · entry tier (Flash-tier models) · one free trial month per person, duplicate trial accounts will be canceled
Unlimited tokens · 1 concurrent generation per stream (extra requests queue — queue waits unbounded, waits can be capped/dropped; 1–50 streams) · 260K context guaranteed
Reachability
Any agent · OpenAI/Anthropic-compatible key
Pricing notes
Meterless: no per-token rates are published, so there is no cost-rule basis. Model ID is "auto" — the model cannot be picked; the fleet rotates (live fleet listed here, the docs page is the authoritative list). The flat price is subsidized by a training/advertising licence over prompts and outputs (no opt-out).
60M input tokens/week · 2x on GLM 5/5.1/5.3, Kimi K2.7 Code, DeepSeek V4 Pro/Pro 0813 · 100 images/day
Reachability
Any agent · OpenAI-compatible key
Pricing notes
Not the best fit for coding agents — large prompts/context burn the weekly cap fast; PAYG safer for heavy coding. Web search is not included in PRO’s subscription coverage.
Cost rule
n/a
MistralProFair usage · up to 6× Free · 150 Flash answers/day
Price confirmed on the provider product page 2026-09-25 (dev.meta.ai/products/muse-code) · High Usage: 5× Everyday usage, more prompts with the latest Muse models
Cost rule
n/a
QwenCloudEssential Plan-37.5% off$10/mo5,625 Credits / 7 days
GPT-6 AstraGPT-6 SolGPT-6 LunaGPT-5.6 SolGPT-5.5GPT-5.6 TerraGPT-5.6 Luna
Ceiling
Unlimited* messages (reasonable use)
Reachability
—
Pricing notes
Codex limits (messages): GPT-6 Luna 350–3,000 = highest · GPT-6 Sol 15–150 · GPT-5.6 Luna 250–2,000 · GPT-5.6 Sol 10–100 · Terra 25–200 · Astra 5–45 · weekly limits may also apply · Luna Reserve = extra Luna-only usage once the regular weekly cap is exhausted (selected Plus accounts, own limit)
Cost rule
n/a
QwenCloudStandard Plan-28% off$18/mo10,000 Credits / 7 days
Grok Bot access is included (not on Lite): link the Grok or X account to grant weekly Grok Bot usage on a Cursor account — cursor.com/help/grok-bot/plans
Price confirmed on the provider product page 2026-09-25 (dev.meta.ai/products/muse-code) · Power Usage: 20× Everyday usage, early access to new features
GPT-6 AstraGPT-6 SolGPT-6 LunaGPT-5.6 Sol ProGPT-5.6 SolGPT-5.6 TerraGPT-5.6 Luna
Ceiling
5× Plus
Reachability
—
Pricing notes
Codex limits (Pro 5×, messages): GPT-6 Luna 1,750–14,000 = highest · GPT-6 Sol 70–700 · GPT-5.6 Luna 1,250–10,000 · GPT-5.6 Sol 50–500 · Terra 125–1,000 · Astra 25–225 · weekly limits may also apply
GPT-6 AstraGPT-6 SolGPT-6 LunaGPT-5.6 Sol ProGPT-5.6 SolGPT-5.6 TerraGPT-5.6 Luna
Ceiling
20× Plus
Reachability
—
Pricing notes
Codex limits (Pro 20×, messages): GPT-6 Luna 7,000–56,000 = highest · GPT-6 Sol 300–3,000 · GPT-5.6 Luna 5,000–40,000 · GPT-5.6 Sol 200–2,000 · Terra 500–4,000 · Astra 100–900 · weekly limits may also apply
Grok Bot access is included; linking SuperGrok Heavy grants the highest linked weekly Grok Bot usage on a Cursor account (cursor.com/help/grok-bot/supergrok) — a usage grant, not a Cursor plan, and no extra usage on top of one.
Cost rule
n/a
No plans match your filters.
Some Subscribe buttons carry referral or affiliate codes — using them may give you a signup bonus or support this project at no extra cost to you.
Markers: ~ proxy estimate, ≈ coarse estimate, and ≥ cache floor. A floor means the plan publishes no cached-read rate, so cached tokens bill at the input rate and the real figure is this or higher. Borrowed cap means the budget is a published API cap at list price, so Req / $1 is a value comparison rather than a measured allowance.
Best-priced plans
Best-priced plans for Kimi K3
The top 5 plans by computed Req/$1 that include Kimi K3, plus the lowest-cost plan. Computed at build from the same cost rules the calculator renders · ~ = proxy estimate
“n/a” = the provider publishes no price structure for Kimi K3 — the calculator renders the same honest n/a. ◆ best value = highest computed Req/$1.
What real work costs
What Kimi K3 costs on real coding tasks
Illustrative cost of common coding tasks on the best-priced plans above, scaled from each plan's live per-request price to a task-typical token volume. Task token assumptions are shared ~proxy estimates — real tasks vary, and they are never per-provider shapes.
Task
CommandCode Go
LLM Gateway DevPass Max
LLM Gateway DevPass Pro
LLM Gateway DevPass Lite
Neural Watt Pro
Quick lookup / one-liner8,000 in · 1,000 out
~1¢
~1¢
~1¢
~1¢
~1¢
Review a 500-line PR60,000 in · 4,000 out
~3¢
~3¢
~3¢
~3¢
~1¢
Fix a bug (agent loop)180,000 in · 12,000 out
~8¢
~8¢
~8¢
~8¢
~3¢
Refactor a module320,000 in · 20,000 out
~13¢
~13¢
~13¢
~13¢
~5¢
Full-repo agent run900,000 in · 45,000 out
~37¢
~37¢
~37¢
~37¢
~15¢
Per-task cost = the plan's effective 77,850-token request price scaled to each task's token volume at the same shape mix. ~ = proxy estimate — shared task-token assumptions, not a quote. “n/a” = Kimi K3 unpriced on that plan.
Common questions
How many requests per $1 do I get on Kimi K3?
The best plan pricing Kimi K3 delivers 49 Req/$1 at the model's typical request shape — on OpenCode Go. Figures are exact confidence: they are computed from published rates and usage caps, proxy estimates are marked with a ~, and borrowed-cap plans use a published API cap at list price, so their Req/$1 is a value comparison rather than a measured allowance.
What's the cheapest plan with Kimi K3?
CommandCode Go at $1/mo. The cheapest plan is not always the best value — check Req/$1 above.
Which plans include Kimi K3?
33 tracked plans across 11 providers price it: CommandCode (Go, GOAT, Max 10×, Max 20×, Pro and Provider), Phoenix Grove (Basic, Pro, Elite, Ultra and Canopy), Standard Compute (Starter, Economy, Standard and Pro), GitHub Copilot (Pro, Pro+ and Max), LLM Gateway DevPass (Lite, Pro and Max), Neural Watt (Basic, Standard and Pro), ZenMux (Builder Starter, Builder Max and Builder Ultra), Ollama (Pro and Max), Ozore (Basic and Pro), Cline Pass and OpenCode Go. 1 more lists it without a published rate, so their calculator rows render n/a: Synthetic Subscription. The calculator compares every tracked monthly plan that prices it.
Where does Kimi K3 rank on the Agent Arena leaderboard?
#8 (Max config) on the Agent Arena agent leaderboard — 6.22% net improvement (board fetched 2026-09-21). Ranks are scraped from https://arena.ai/leaderboard/agent, never typed by hand.
What is Kimi K3?
Kimi K3 is Moonshot AI’s flagship model for agentic coding and reasoning. It is a proprietary model (not open-weight), reachable only through provider endpoints.
Can Kimi K3 be used with any agent (BYOK)?
Yes — several providers expose Kimi K3 through an OpenAI-compatible endpoint with no CLI lock, so you can use it with Claude Code, Cursor, OpenCode, or any agent that supports custom endpoints. Some plans are tool-locked to a specific CLI, but there are plenty of open BYOK options.
What kind of model is Kimi K3?
Kimi K3 is a proprietary text generation model (family: moonshot, kind: text) focused on agentic coding and reasoning.
What is Kimi K3 best at?
Kimi K3 is suited to multi-step agent tasks, coding workflows, and problems that benefit from extended reasoning.
Newsletter
The weekly brief — plan churn, dated.
A plain weekly digest of every tracked price move, ceiling change and model-list update — pulled straight from the dated changelog ledger, with a one-line “worth switching?” note on each. Free. No spam, unsubscribe anytime.