Kimi k2.6:thinking API
Moonshot's Kimi k2.6:thinking costs $0.501 per million input tokens and $2.6 per million output tokens from 1 supplier. Each request through the OpenAI-compatible API goes to the offer with the lowest estimated charge that meets your rules.
- Providers compete for each request. The exchange ranks the offers that meet your rules by estimated charge. It fixes your maximum charge before the first token, and you pay for the tokens that the model uses.
- The OpenAI and Anthropic APIs. Use the OpenAI or Anthropic SDK that you have. Set the base URL and your key, and send kimi-k2.6:thinking as the model.
- Your data rules. Send x-liquid-retention: none and only offers whose provider does not keep your content compete for the request.
About Kimi k2.6:thinking
Kimi k2.6:thinking is a language model from Moonshot. Its release date is 20 April 2026. Multimodal Kimi workhorse for agent loops, coding tasks, and visual context. Its weights are open. Its training data ends in January 2025. It takes text and images and gives text. The largest offer takes 262,144 input tokens. It has 1 supplier on this exchange now.
Description, release date and licence: models.dev (https://models.dev/)
Facts
- Model name
- kimi-k2.6:thinking
- Maker
- Moonshot
- Context window
- 262.1K tokens
- Input price
- $0.501 per million tokens
- Output price
- $2.6 per million tokens
- Suppliers
- 1 supplier
- Regions
- us
Providers
- Architect: us, declared precision provider-managed, $0.501 input and $2.6 output per million tokens, $0.125 cache read per million tokens, 262.1K tokens of context
Scores
- HLE: 37.5, #78 of 465
- GPQA: 91.1, #45 of 432
- AA-LCR: 81.0, #44 of 409
- SciCode: 51.5, #59 of 178
- Terminal-Bench Hard: 43.9, #26 of 311
- τ²-Bench: 95.9, #18 of 313
- IFBench: 76.0, #23 of 321
- SimpleQA Verified: 34.9, #73 of 172
Scores: Artificial Analysis (https://artificialanalysis.ai/)
Scores: Epoch AI, 'Capabilities & Benchmarking' (https://epoch.ai/benchmarks)
Request
curl https://router.inference.ai.exchange/v1/chat/completions \
-H "Authorization: Bearer $LIQUID_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "kimi-k2.6:thinking",
"messages": [{"role":"user","content":"Say hello."}]
}'Questions
- What is Kimi k2.6:thinking?
- Kimi k2.6:thinking is a language model from Moonshot. Its release date is 20 April 2026. Multimodal Kimi workhorse for agent loops, coding tasks, and visual context. Its weights are open. Its training data ends in January 2025.
- How much does Kimi k2.6:thinking cost?
- Offers now ask $0.501 per million input tokens and $2.6 per million output tokens. A job pays for the tokens it uses. The exchange picks the cheapest eligible offer.
- What is the context length of Kimi k2.6:thinking?
- The largest offer takes 262,144 input tokens. A smaller offer can take less. The exchange skips an offer that cannot take your message.
- What can Kimi k2.6:thinking do?
- Its providers list Vision, Reasoning effort, JSON object, JSON schema, Frequency penalty, Logit bias, Log probabilities, Min-p sampling, Presence penalty, Repetition penalty, Seed, Stop sequences, Top-k sampling, Top-p sampling, Text, Tools.
- What inputs and outputs does Kimi k2.6:thinking support?
- Text · Images → Text.
- Which providers serve Kimi k2.6:thinking?
- It has 1 supplier now, with 1 offer. The table above ranks them by the cost of one agent-mix job.
- How do I call Kimi k2.6:thinking?
- Send "kimi-k2.6:thinking" as the model in an OpenAI or Anthropic request. The API section links to setup steps for each client.
- When was Kimi k2.6:thinking released?
- Its creator released it on 20 April 2026.
- When was Kimi k2.6:thinking listed?
- The exchange first listed it on 10/10/2026. This is not its release date.
Related pages
- Kimi K3
- Kimi K2 Instruct
- Kimi K2 Instruct 0711
- Kimi K2 Instruct FP4
- Kimi K2 Instruct 0905
- Kimi K2 Instruct Fast
- Popular open weights
- Reasoning models
- Set up a client
- All models
- Providers
- Discover the best model for your budget
- Quickstart
- llms.txt: the index of every page, as Markdown
- llms-full.txt: every page in one file
- This page as Markdown