GPT-5.4
Input: $2.5 / million tokens
Output: $15 / million tokens
Cache read: $0.25 / million tokens
Context: 1050000
截至 2026-05，按公开价格页样例 $4.50/$30.00 每百万 tokens，汇率 7.2 折算；请以上线时官方价格页为准。
Usage based API
Billed by usage; no assumed monthly cost
Understand the fit and limits
Use cases
Open-source coding agents with free models, BYOK, local models or Gateway. Compare actual usage and governance needs.
Keep in mind
Free software is not unlimited paid inference. A 5% fee applies to top-ups and Pass purchases/renewals. Mainland access, payment and official CNY prices are unverified. Conflicting consumption-order sources prevent an exact discounted total.
Entitlements and usage
- Allowance
- Inference billed by actual tokens; credit purchase processing fees are separate.
- Benefits
- Usage-based inference at provider list prices; credit purchases add a separate 5% processing fee.
- Models listed in source
- Anthropic: Claude Sonnet 4.6, OpenAI: GPT-5.4, Qwen: Qwen3 Coder Next, Auto Free
- Source notes and limits
- Usage-based inference at provider list prices; credit purchases add a separate 5% processing fee.
API input and output pricing
Claude Sonnet 4.6
Input: $3 / million tokens
Output: $15 / million tokens
Cache read: $0.3 / million tokens
Cache write: $3.75 / million tokens
Context: 1000000
截至 2026-05，示例占位型号；价格按 Sonnet 档公开美元价 $3/$15 每百万 tokens，汇率 7.2 折算。
qwen3-coder-next
Input: $0.3 / million tokens
Output: $1.5 / million tokens
Supported models
Costs and credit rules
Free software does not include paid-model inference. Seats, inference and cloud compute are billed separately. Do not add inference already paid with credits again. BYOK provider invoices are separate from platform charges.
Each $1 of base credits funds $1 of usage. Processing fee: 5%, without additional credits. Applies to: credit top-ups, credit subscription purchases, credit subscription renewals.
Current Sep 30, 2026 terms limit purchased credits to one year or earlier account deletion. New users are covered immediately; existing users at the earliest acceptance, Desktop/artifact use or 30 days later. A billing help subsection still says never expire; check the applicable terms.
The FAQ says earliest-expiring credits first; terms say promotional, then monthly/purchased, then bonus. These conflict, so exact consumption-order estimates are disabled.
- Free software does not include paid inference. Configure primary, autocomplete and background models separately for free usage.
- Credit top-ups and Pass purchase/renewal add a separate 5% processing fee without extra credits; Teams seats do not include credits.
- Provider keys and existing Coding Plans are billed by their provider. Do not add inference already paid with credits a second time.
- Cloud compute and inference are separate, billed per second; the $5 launch balance is not an extra charge.
- Canceling renewals resets the streak. Annual Pass base credits continue to be issued monthly for the paid term.
Organization Pass capacity is purchased per paid seat, pooled and adjusted with seat count. Personal subscriptions remain separate.
- Code Review: $0.33 / hour · Billed per second; inference separate
- Cloud Agent — Docker: $0.6 / hour · Billed per second; inference separate
- Cloud Agent — Small: $0.6 / hour · Billed per second; inference separate
- Cloud Agent — Standard: $1.2 / hour · Billed per second; inference separate
International product. Official mainland-China access, account and payment eligibility and CNY pricing are not established. CN lists free-tool information only; BytePlus support does not verify Volcengine China Coding Plan compatibility.
BYOK and existing subscriptions
- Own provider API key: Anthropic, AWS Bedrock, Azure Foundry (experimental), DeepSeek, Fireworks, Google AI Studio, Inception, MiniMax, Mistral AI, Moonshot AI, Nebius Token Factory, Novita, OpenAI, Xiaomi, SpaceXAI, Z.ai
- Verified Coding Plans: BytePlus Coding Plan, Kimi Code, Z.ai Coding Plan, OpenCode Go, Xiaomi Token Plan (Europe), Xiaomi Token Plan (Singapore)
- Third-party routing: OpenRouter, Requesty, Ollama Cloud
- Local models: Ollama, LM Studio, OpenAI-compatible endpoint
- Compatibility unverified: Volcengine China Coding Plan
Provider account, plan-specific keys, endpoints, models and usage restrictions still apply. API-key support does not make all subscriptions interchangeable. An invalid Gateway BYOK key does not fall back to a paid platform key. Official BYOK setup
Selected Gateway prices (USD / million tokens)
| Model | Official model ID | Input | Output | Context tokens | Conditions |
|---|---|---|---|---|---|
| Anthropic: Claude Sonnet 4.6 | anthropic/claude-sonnet-4.6 | $3 | $15 | 1000000 | Billed by actual tokens |
| OpenAI: GPT-5.4 | openai/gpt-5.4 | $2.5 | $15 | 1050000 | Billed by actual tokens |
| Qwen: Qwen3 Coder Next | qwen/qwen3-coder-next | $0.12 | $0.8 | 262144 | Billed by actual tokens |
| Auto Free | kilo-auto/free | $0 | $0 | 256000 | Dynamic free route; availability and limits change. Prompts may be logged for training. |
Official prices and rules · 2026-10-12T01:56:59.111Z
Choose a purchase channel
Kilo Code Kilo Gateway · Individual plan · Usage based
Official websiteBrand direct
Usage based API
Monthly equivalent; full period paid upfront
Activation: check official terms
No verified quotes for this billing period or scope
View source and termsPayment is completed with the provider. Final prices, eligibility and after-sales terms are set by the purchase page.
Sources and verification
Last fact check: 2026-10-12
Data generation is not a fact check. If information differs, check the latest official source.
Frequently asked questions
- How much does Kilo Code Kilo Gateway cost?
- Kilo Code Kilo Gateway is priced at Usage based API · Monthly equivalent; full period paid upfront.
- Which models does Kilo Code Kilo Gateway support?
- Representative models: Anthropic: Claude Sonnet 4.6, OpenAI: GPT-5.4, Qwen: Qwen3 Coder Next, Auto Free.
- When was Kilo Code Kilo Gateway's price last verified?
- CreditsPlan last verified this plan on 2026-10-12 against kilo.ai; the vendor's official site prevails.