Kimi K3 pricing: real cost per million tokens
What Kimi's Kimi K3 really costs across 8 plans from 5 providers: monthly fees, usable allowances and real per-token prices, against the official API list price and independent leaderboards.
Data updated · How real prices are computed
As of 2026-10-08, the cheapest way to use Kimi K3 is Kimi Allegro at a real cost of about $0.02274 per million tokens ($99 (¥699)/month for roughly 4.353B tokens a month). We track 8 plans and API channels that serve Kimi K3. Official API list price per million tokens: $0.3 cached input, $3 input, $15 output — about $0.441 per million tokens on a standard workload mix.
- Lowest real price
- $0.02274/MTokKimi Allegro
- API list price
- $0.441/MTokstandard workload mix
- Most monthly tokens
- 4.353BKimi Allegro
- Plans & APIs
- 88 subscriptions · 0 metered APIs
Plot Kimi K3 on the price–capability chart →
Kimi K3 plans and real prices
Real price = monthly fee ÷ tokens you can actually use in a month. “vs API” shows how many times cheaper than the official API list price (standard workload mix) the plan is. “Evidence” opens the full calculation and sources for that row.
| Plan | Provider | Billing | Monthly fee | Monthly tokens | Real $/MTok | vs API | Confidence | Data date | |
|---|---|---|---|---|---|---|---|---|---|
| Kimi Allegro | Kimi | Subscription | $99 (¥699)/mo | 4.353B | $0.02274 | 19.4× | Medium | Evidence ↓ | |
| Kimi Allegretto | Kimi | Subscription | $39 (¥199)/mo | 1.451B | $0.02688 | 16.4× | Medium | Evidence ↓ | |
| Droid Max | Factory | Subscription | $200/mo | 3.368B | $0.05938 | 7.4× | Medium | Evidence ↓ | |
| Kimi Moderato | Kimi | Subscription | $19 (¥99)/mo | 0.29B | $0.06552 | 6.7× | Medium | Evidence ↓ | |
| Ollama Pro | Ollama | Subscription | $20/mo | 0.136B | $0.147 | 3× | Medium | Evidence ↓ | |
| Ollama Max | Ollama | Subscription | $100/mo | 0.68B | $0.147 | 3× | Medium | Evidence ↓ | |
| Command Code GOAT | Command Code | Subscription | $10/mo | 0.045B | $0.2203 | 2× | Medium | Evidence ↓ | |
| OpenCode Go | OpenCode | Subscription | $10/mo | 0.034B | $0.2941 | 1.5× | Medium | Evidence ↓ |
Behind each number
For every plan: the data date, the exact calculation, the quota basis, how leaderboard scores were mapped, and the original adoption evidence with its sources.
Kimi AllegroKimi$0.0227429/MTok
- Plan
- Kimi Allegro · Kimi
- Data date
- 2026-07-25 – 2026-07-26 · derived; uses the data date of Kimi 会员 199 · kimi-k3
- Real price / MTok
- $0.0227429
- Monthly tokens
- 4,353,000,000 (4.353B)
- Quota confidence
- Medium
- Calculation
- $99 / month (¥699) ÷ 4,353,000,000 tokens × 1,000,000 ≈ $0.0227429 / MTok
- Quota basis
Not workload-normalized: sample lacks a token breakdown (or is an official absolute token table); raw tokens used as-is
Official API list (cached / in / out, per MTok): $0.3 / $3 / $15 · standard-load blended ≈ $0.441/MTok
- Benchmark mapping
- Artificial Analysis Intelligence Index v4.3: Medium (2 configurations)
Terminal-Bench 4.0 (AA): Low (2 configurations)
Code Arena | WebDev Overall: Low (1 configuration)
Agent Arena · Overall (Ranking / Pareto): Low (1 configuration)
Artificial Analysis Coding Agent Index v1.5: Low (1 configuration)
OpenDesign Arena · Quality ranking: Low (1 configuration)
DeepSWE v1.1: Low (1 configuration)Mapping confidence: how directly each score refers to the model served here.
Adoption evidence · source and rationale (translated from the Chinese record)
¥199 tier × official 60/20; kimi-adoption-round6-2026-09-08.json
Old 3.483B→4.353B: follows the ¥199 tier's switch to weekly pool ×5; inherits the K3-256K-dominated mixed-workload estimate, not a pure K3 1M model measurement; same-name domestic and overseas tiers merged into one point: monthly fee and unit price use the international Kimi Allegro $99 list price (previously domestic ¥699÷6.7787≈$103.12); price/currency still record the actual domestic price paid; the allowance stays on the domestic-tier basis — the overseas same-name tier's absolute tokens are unmeasured; the merge is for price-tier display only
kimi-adoption-round6-2026-09-08.jsonBrowse the public evidence archive
Kimi AllegrettoKimi$0.026878/MTok
- Plan
- Kimi Allegretto · Kimi
- Data date
- 2026-07-25 – 2026-07-26 · sampled
- Real price / MTok
- $0.026878
- Monthly tokens
- 1,451,000,000 (1.451B)
- Quota confidence
- Medium
- Calculation
- $39 / month (¥199) ÷ 1,451,000,000 tokens × 1,000,000 ≈ $0.026878 / MTok
- Quota basis
Not workload-normalized: sample lacks a token breakdown (or is an official absolute token table); raw tokens used as-is
Official API list (cached / in / out, per MTok): $0.3 / $3 / $15 · standard-load blended ≈ $0.441/MTok
- Benchmark mapping
- Artificial Analysis Intelligence Index v4.3: Medium (2 configurations)
Terminal-Bench 4.0 (AA): Low (2 configurations)
Code Arena | WebDev Overall: Low (1 configuration)
Agent Arena · Overall (Ranking / Pareto): Low (1 configuration)
Artificial Analysis Coding Agent Index v1.5: Low (1 configuration)
OpenDesign Arena · Quality ranking: Low (1 configuration)
DeepSWE v1.1: Low (1 configuration)Mapping confidence: how directly each score refers to the model served here.
Adoption evidence · source and rationale (translated from the Chinese record)
Local ccusage 243,739,068/84% back-derives the weekly quota × the Kimi monthly pool's 5×; kimi-adoption-round6-2026-09-08.json
Old 1.161B→1.451B: confirmed Kimi's monthly pool = weekly pool ×5 — the old value wrongly applied the project-wide 4 weeks; the sample is dominated by k3-256k and includes kimi-for-coding — not a pure K3 1M measurement; the SWE 1.7 short-window panel has a different model/statistics window and did not replace the baseline; ACP 14.28 is old-model corroboration, not directly adopted; not workload-normalized: ccusage stores only the weekly total, no breakdown; same-name domestic and overseas tiers merged into one point: monthly fee and unit price use the international Kimi Allegretto $39 list price (previously domestic ¥199÷6.7787≈$29.36); price/currency still record the actual domestic price paid; the allowance stays on the domestic-tier basis — the overseas same-name tier's absolute tokens are unmeasured; the merge is for price-tier display only
kimi-adoption-round6-2026-09-08.jsonBrowse the public evidence archive
Droid MaxFactory$0.0593824/MTok
- Data date
- 2026-09-29 · derived; uses the data date of Droid Max · claude-opus-5.5
- Real price / MTok
- $0.0593824
- Monthly tokens
- 3,368,000,000 (3.368B)
- Quota confidence
- Medium
- Calculation
- $200 / month ÷ 3,368,000,000 tokens × 1,000,000 ≈ $0.0593824 / MTok
- Quota basis
Anthropic workload: 97% cache reads / 2.5% cache writes / 0.5% output
Official API list (cached / in / out, per MTok): $0.3 / $3 / $15 · standard-load blended ≈ $0.441/MTok
- Benchmark mapping
- Artificial Analysis Intelligence Index v4.3: Medium (2 configurations)
Terminal-Bench 4.0 (AA): Low (2 configurations)
Code Arena | WebDev Overall: Low (1 configuration)
Agent Arena · Overall (Ranking / Pareto): Low (1 configuration)
Artificial Analysis Coding Agent Index v1.5: Low (1 configuration)
OpenDesign Arena · Quality ranking: Low (1 configuration)
DeepSWE v1.1: Low (1 configuration)Mapping confidence: how directly each score refers to the model served here.
Adoption evidence · source and rationale (translated from the Chinese record)
Derived from same-plan claude-opus-5.5 2.526B × 1.3333333333333335
Factory official multipliers Opus 5.5 1.6× / kimi-k3 1.2×: the same Standard Usage pool scaled by multiplier — not a measurement of this model; inherits the Opus 5.5 row's Anthropic-tier workload, not recomputed for each model's own price gap; docs.factory.ai/docs/models; droid-model-multipliers-2026-09-29.json; Droid Core models draw Standard Usage first, then a separate free open-source pool applies (independent cap unmeasured); this value counts only the Standard Usage part
droid-model-multipliers-2026-09-29.jsonBrowse the public evidence archive
Kimi ModeratoKimi$0.0655172/MTok
- Plan
- Kimi Moderato · Kimi
- Data date
- 2026-07-25 – 2026-07-26 · derived; uses the data date of Kimi 会员 199 · kimi-k3
- Real price / MTok
- $0.0655172
- Monthly tokens
- 290,000,000 (0.29B)
- Quota confidence
- Medium
- Calculation
- $19 / month (¥99) ÷ 290,000,000 tokens × 1,000,000 ≈ $0.0655172 / MTok
- Quota basis
Not workload-normalized: sample lacks a token breakdown (or is an official absolute token table); raw tokens used as-is
Official API list (cached / in / out, per MTok): $0.3 / $3 / $15 · standard-load blended ≈ $0.441/MTok
- Benchmark mapping
- Artificial Analysis Intelligence Index v4.3: Medium (2 configurations)
Terminal-Bench 4.0 (AA): Low (2 configurations)
Code Arena | WebDev Overall: Low (1 configuration)
Agent Arena · Overall (Ranking / Pareto): Low (1 configuration)
Artificial Analysis Coding Agent Index v1.5: Low (1 configuration)
OpenDesign Arena · Quality ranking: Low (1 configuration)
DeepSWE v1.1: Low (1 configuration)Mapping confidence: how directly each score refers to the model served here.
Adoption evidence · source and rationale (translated from the Chinese record)
¥199 tier × official 4/20; kimi-adoption-round6-2026-09-08.json
Old 232M→290M: follows the ¥199 tier's switch to weekly pool ×5; inherits the K3-256K-dominated mixed-workload estimate, not a pure K3 1M model measurement; same-name domestic and overseas tiers merged into one point: monthly fee and unit price use the international Kimi Moderato $19 list price (previously domestic ¥99÷6.7787≈$14.60); price/currency still record the actual domestic price paid; the allowance stays on the domestic-tier basis — the overseas same-name tier's absolute tokens are unmeasured; the merge is for price-tier display only
kimi-adoption-round6-2026-09-08.jsonBrowse the public evidence archive
Ollama ProOllama$0.146951/MTok
- Plan
- Ollama Pro · Ollama
- Data date
- 2026-09-06 · official source date
- Real price / MTok
- $0.146951
- Monthly tokens
- 136,100,000 (0.136B)
- Quota confidence
- Medium
- Calculation
- $20 / month ÷ 136,100,000 tokens × 1,000,000 ≈ $0.146951 / MTok
- Quota basis
Standard workload: 97% cache reads / 2.5% input / 0.5% output
Official API list (cached / in / out, per MTok): $0.3 / $3 / $15 · standard-load blended ≈ $0.441/MTok
- Benchmark mapping
- Artificial Analysis Intelligence Index v4.3: Medium (2 configurations)
Terminal-Bench 4.0 (AA): Low (2 configurations)
Code Arena | WebDev Overall: Low (1 configuration)
Agent Arena · Overall (Ranking / Pareto): Low (1 configuration)
Artificial Analysis Coding Agent Index v1.5: Low (1 configuration)
OpenDesign Arena · Quality ranking: Low (1 configuration)
DeepSWE v1.1: Low (1 configuration)Mapping confidence: how directly each score refers to the model served here.
Adoption evidence · source and rationale (translated from the Chinese record)
https://ollama.com/pricing official usage credits and three-part pricing; https://ollama.com/blog/transparent-pricing; code-subscriptions-round1-2026-09-06.json; ollama-deepseek-v41-round1-2026-09-11.json
Added 136.1M: shared monthly pool $60 ÷ project standard-workload weighted price; official three-part pricing. Allowances of different models in the same plan are not additive (assuming the shared pool is used up on a single model); no usage-panel token+% screenshot — computed from official absolute credits and the price list
https://ollama.com/pricinghttps://ollama.com/blog/transparent-pricingcode-subscriptions-round1-2026-09-06.jsonollama-deepseek-v41-round1-2026-09-11.jsonBrowse the public evidence archive
Ollama MaxOllama$0.146994/MTok
- Plan
- Ollama Max · Ollama
- Data date
- 2026-09-06 · official source date
- Real price / MTok
- $0.146994
- Monthly tokens
- 680,300,000 (0.68B)
- Quota confidence
- Medium
- Calculation
- $100 / month ÷ 680,300,000 tokens × 1,000,000 ≈ $0.146994 / MTok
- Quota basis
Standard workload: 97% cache reads / 2.5% input / 0.5% output
Official API list (cached / in / out, per MTok): $0.3 / $3 / $15 · standard-load blended ≈ $0.441/MTok
- Benchmark mapping
- Artificial Analysis Intelligence Index v4.3: Medium (2 configurations)
Terminal-Bench 4.0 (AA): Low (2 configurations)
Code Arena | WebDev Overall: Low (1 configuration)
Agent Arena · Overall (Ranking / Pareto): Low (1 configuration)
Artificial Analysis Coding Agent Index v1.5: Low (1 configuration)
OpenDesign Arena · Quality ranking: Low (1 configuration)
DeepSWE v1.1: Low (1 configuration)Mapping confidence: how directly each score refers to the model served here.
Adoption evidence · source and rationale (translated from the Chinese record)
https://ollama.com/pricing official usage credits and three-part pricing; https://ollama.com/blog/transparent-pricing; code-subscriptions-round1-2026-09-06.json; ollama-deepseek-v41-round1-2026-09-11.json
Added 680.3M: shared monthly pool $300 ÷ project standard-workload weighted price; official three-part pricing. Allowances of different models in the same plan are not additive (assuming the shared pool is used up on a single model); no usage-panel token+% screenshot — computed from official absolute credits and the price list
https://ollama.com/pricinghttps://ollama.com/blog/transparent-pricingcode-subscriptions-round1-2026-09-06.jsonollama-deepseek-v41-round1-2026-09-11.jsonBrowse the public evidence archive
Command Code GOATCommand Code$0.220264/MTok
- Data date
- 2026-09-30 · official source date
- Real price / MTok
- $0.220264
- Monthly tokens
- 45,400,000 (0.045B)
- Quota confidence
- Medium
- Calculation
- $10 / month ÷ 45,400,000 tokens × 1,000,000 ≈ $0.220264 / MTok
- Quota basis
Standard workload: 97% cache reads / 2.5% input / 0.5% output
Official API list (cached / in / out, per MTok): $0.3 / $3 / $15 · standard-load blended ≈ $0.441/MTok
- Benchmark mapping
- Artificial Analysis Intelligence Index v4.3: Medium (2 configurations)
Terminal-Bench 4.0 (AA): Low (2 configurations)
Code Arena | WebDev Overall: Low (1 configuration)
Agent Arena · Overall (Ranking / Pareto): Low (1 configuration)
Artificial Analysis Coding Agent Index v1.5: Low (1 configuration)
OpenDesign Arena · Quality ranking: Low (1 configuration)
DeepSWE v1.1: Low (1 configuration)Mapping confidence: how directly each score refers to the model served here.
Adoption evidence · source and rationale (translated from the Chinese record)
https://commandcode.ai/docs/plans/goat official per-model allowance and three-part pricing; https://commandcode.ai/pricing; $10->$70 credits; code-subscriptions-round1-2026-09-06.json; goat-opencode-catalogs-round3-2026-09-30.json
Added 45.4M: min(shared monthly pool $70, model allowance $20) ÷ project standard-workload weighted price; official three-part price; New models table $20 (temporary $60 promo through 10/7, not adopted per temp-promo policy). Processing fee ignored; allowances of different models in the same plan are not additive; no usage-panel token+% screenshot — computed from official absolute credits and the price list
https://commandcode.ai/docs/plans/goathttps://commandcode.ai/pricing;code-subscriptions-round1-2026-09-06.jsongoat-opencode-catalogs-round3-2026-09-30.jsonBrowse the public evidence archive
OpenCode GoOpenCode$0.294118/MTok
- Plan
- OpenCode Go · OpenCode
- Data date
- 2026-09-30 · official source date
- Real price / MTok
- $0.294118
- Monthly tokens
- 34,000,000 (0.034B)
- Quota confidence
- Medium
- Calculation
- $10 / month ÷ 34,000,000 tokens × 1,000,000 ≈ $0.294118 / MTok
- Quota basis
Standard workload: 97% cache reads / 2.5% input / 0.5% output
Official API list (cached / in / out, per MTok): $0.3 / $3 / $15 · standard-load blended ≈ $0.441/MTok
- Benchmark mapping
- Artificial Analysis Intelligence Index v4.3: Medium (2 configurations)
Terminal-Bench 4.0 (AA): Low (2 configurations)
Code Arena | WebDev Overall: Low (1 configuration)
Agent Arena · Overall (Ranking / Pareto): Low (1 configuration)
Artificial Analysis Coding Agent Index v1.5: Low (1 configuration)
OpenDesign Arena · Quality ranking: Low (1 configuration)
DeepSWE v1.1: Low (1 configuration)Mapping confidence: how directly each score refers to the model served here.
Adoption evidence · source and rationale (translated from the Chinese record)
https://opencode.ai/docs/go/ official per-model Usage and three-part pricing; opencode-go-round5-2026-09-06.json; goat-opencode-catalogs-round3-2026-09-30.json
Old: 38.1M (request estimate)→34M: min(shared monthly pool $60, model Usage $15) ÷ project standard-workload weighted price; single official tier. The official request counts serve only as a cross-check and are no longer the primary allowance figure; allowances of different models in the same plan are not additive
https://opencode.ai/docs/go/goat-opencode-catalogs-round3-2026-09-30.jsonopencode-go-round5-2026-09-06.jsonBrowse the public evidence archive
Kimi K3 benchmark scores
Highest archived configuration per leaderboard; ranks are among all models each board archives in this dataset. Scores are never mixed across boards.
| Leaderboard | Metric | Score | Rank | Configuration | Effort | Harness | Configs | Snapshot |
|---|---|---|---|---|---|---|---|---|
| Artificial Analysis Intelligence Index v4.3 | Intelligence Index | 43.59 | #17 / 60 | Kimi K3 (max) | Max | — | 2 | 2026-10-01 |
| Terminal-Bench 4.0 (AA) | Resolution Rate % | 12.63 | #28 / 53 | Artificial Analysis - Kimi K3 (low) | Low | Artificial Analysis | 2 | 2026-10-01 |
| Code Arena | WebDev Overall | Arena Score | 1,658 | #10 / 54 | kimi-k3-max | Max | — | 1 | 2026-09-30 |
| Agent Arena · Overall (Ranking / Pareto) | Net Improvement % | 3.98 | #14 / 39 | Kimi K3 (Max) | Max | — | 1 | 2026-09-30 |
| Artificial Analysis Coding Agent Index v1.5 | Coding Agent Index | 51.93 | #12 / 19 | Kimi Code CLI - Kimi K3 | — | Kimi Code CLI | 1 | 2026-10-01 |
| OpenDesign Arena · Quality ranking | Average task score | 65.7 | #16 / 16 | Kimi K3 | — | OpenDesign | 1 | 2026-10-01 |
| DeepSWE v1.1 | Pass@1 % | 68.51 | #11 / 33 | mini-swe-agent - kimi-k3 (max) | Max | mini-swe-agent | 1 | 2026-09-03 |
Every archived configuration
Artificial Analysis Intelligence Index v4.3 2 configurations
| Configuration | Score | Score interval | Harness / effort / mode | Source task cost (mean / median) | Checked | Links |
|---|---|---|---|---|---|---|
| Kimi K3 (max) | 43.5938 | — | — / Max / — | — / — | 2026-10-01 | Source · Archive |
| Kimi K3 (low) | 30.0667 | — | — / Low / — | — / — | 2026-10-01 | Source · Archive |
Terminal-Bench 4.0 (AA) 2 configurations
| Configuration | Score | Score interval | Harness / effort / mode | Source task cost (mean / median) | Checked | Links |
|---|---|---|---|---|---|---|
| Artificial Analysis - Kimi K3 (low)AA's own-harness independent measurement of Terminal-Bench 4.0 (the terminalBench40 field in the models-board payload, fraction×100); scored separately from the official tbench.ai board. | 12.6263 | — | Artificial Analysis / Low / — | — / — | 2026-10-01 | Source · Archive |
| Artificial Analysis - Kimi K3 (max)AA's own-harness independent measurement of Terminal-Bench 4.0 (the terminalBench40 field in the models-board payload, fraction×100); scored separately from the official tbench.ai board. | 12.6263 | — | Artificial Analysis / Max / — | — / — | 2026-10-01 | Source · Archive |
Code Arena | WebDev Overall 1 configuration
Agent Arena · Overall (Ranking / Pareto) 1 configuration
Artificial Analysis Coding Agent Index v1.5 1 configuration
| Configuration | Score | Score interval | Harness / effort / mode | Source task cost (mean / median) | Checked | Links |
|---|---|---|---|---|---|---|
| Kimi Code CLI - Kimi K3Score belongs to this named agent/harness × model configuration; it is not a harness-independent model measurement. indexScore stored as percent (raw fraction × 100). | 51.9259 | — | Kimi Code CLI / — / — | $5.047 / — | 2026-10-01 | Source · Archive |
OpenDesign Arena · Quality ranking 1 configuration
DeepSWE v1.1 1 configuration
These are reference mappings, not benchmarks of a subscription or API channel. Harness and quota-measurement effort alignment are unverified. Source task costs are not subscription task costs.
- Exact served-model reference; quota-measurement effort is unverified.
- Exact served-model reference only; product harness and quota-measurement effort are unverified. Not a benchmark measurement of this subscription or API channel.
Kimi K3 official API pricing
| Cached input | Input | Output |
|---|---|---|
| $0.3 | $3 | $15 |
Frequently asked questions
How much does Kimi K3 really cost per million tokens?
Measured as monthly fee ÷ tokens you can actually use, Kimi K3 ranges from $0.02274 (Kimi Allegro) to $0.2941 (OpenCode Go) per million tokens.
Which subscription gives the most Kimi K3 usage?
Kimi Allegro allows roughly 4.353B tokens of Kimi K3 per month for $99 (¥699).
What is the official API price of Kimi K3?
Official API list price per million tokens: $0.3 cached input, $3 input, $15 output — about $0.441 per million tokens on a standard workload mix.
How does Kimi K3 score on independent leaderboards?
Kimi K3 has scores on 7 leaderboards; its strongest placement is 1,658 (Arena Score) on Code Arena | WebDev Overall, #10 of 54 models tracked.
Who makes Kimi K3, and where can I use it?
Kimi K3 (model ID kimi-k3) is made by Kimi and is available through Command Code, Factory, Kimi, Ollama, OpenCode.
More Kimi models
| Model | Lowest real $/MTok | Cheapest plan | Plans |
|---|---|---|---|
| Kimi K2.7 Code | $0.02105 | Kimi Allegro | 8 |
| Kimi K2.6 | $0.03316 | OpenCode Go | 2 |
| Kimi K2.5 | $0.06349 | Command Code GOAT | 1 |
| Kimi K2.7 Code HighSpeed | $0.2278 | Command Code GOAT | 1 |