HIVE
Home / _core / MODELS-MAP — OpenRouter Model Catalog

MODELS-MAP — OpenRouter Model Catalog

Edit on GitHub
On this page

Source of truth for skills that call models via the OpenRouter MCP. Zero defaults, zero prescription. Skills always ask which model to use. Last updated: 2026-05-31

#How to use

In skills: always ask the model via AskUserQuestion if not passed as argument. Accepts 3 formats:

  1. Friendly alias (e.g. kimi, gpt5.5, grok) — resolved against the table below
  2. Raw OpenRouter ID (e.g. moonshotai/kimi-k2.6, openai/gpt-5.4-pro)
  3. OpenRouter auto-update aliases (e.g. ~google/gemini-pro-latest) — always points to latest version

Example invocations:

/second-opinion                          # asks for model
/second-opinion kimi "my question"       # alias
/second-opinion moonshotai/kimi-k2.6 "..." # raw ID
/consensus [kimi,gpt5.5,grok] "..."      # list

#💻 Coding (ranked by LiveCodeBench v6 / SWE-Bench Pro — May 2026)

Alias OpenRouter ID Ctx $/M in $/M out Benchmark
kimi moonshotai/kimi-k2.6 256k $0.68 $3.42 #1 LiveCodeBench v6: 89.6% · SWE-Pro leader
kimi-think moonshotai/kimi-k2-thinking 256k $0.60 $2.50 Reasoning + coding combined
gpt5codex openai/gpt-5.3-codex 400k $1.75 $14.00 OpenAI dedicated coding model
gpt5codex-max openai/gpt-5.1-codex-max 400k $1.25 $10.00 Codex large variant
qwencoder qwen/qwen3-coder-plus 1M $0.65 $3.25 87.1% LiveCodeBench v6
qwencoder-next qwen/qwen3-coder-next 256k $0.11 $0.80 Efficient per active param
glm5.1 z-ai/glm-5.1 200k $0.98 $3.08 #1 web dev open (1534 Elo)
codestral mistralai/codestral-2508 256k $0.30 $0.90 European coding, low cost

#🧠 Reasoning / Flagship

Alias OpenRouter ID Ctx $/M in $/M out Benchmark
gpt5.5 openai/gpt-5.5 1M $5.00 $30.00 82.7% Terminal-Bench 2.0 (top agentic)
gpt5.4 openai/gpt-5.4 1M $2.50 $15.00 #1 BenchLM weighted: 73.9
grok x-ai/grok-4.20 2M $1.25 $2.50 Largest viable ctx + low cost
minimax minimax/minimax-m2.7 200k $0.26 $1.20 MiniMax reasoning tier
qwen-max qwen/qwen3.7-max 1M $1.25 $3.75 Top Qwen flagship
qwen-think qwen/qwen3-max-thinking 256k $0.78 $3.90 Qwen with chain-of-thought

#⚡ Fast / cheap (recurring workers, atomic tasks)

Alias OpenRouter ID Ctx $/M in $/M out
dsflash deepseek/deepseek-v4-flash 1M $0.10 $0.20
qwenflash qwen/qwen3.6-flash 1M $0.19 $1.13
flashlite google/gemini-3.1-flash-lite 1M $0.25 $1.50
glm-flash z-ai/glm-4.7-flash 200k $0.06 $0.40
granite ibm-granite/granite-4.1-8b 128k $0.05 $0.10
gpt5.4-nano openai/gpt-5.4-nano 400k $0.20 $1.25

#🌍 Long context / multimodal

Alias OpenRouter ID Ctx $/M in $/M out
gemini-pro ~google/gemini-pro-latest 1M $2.00 $12.00
gemini-flash ~google/gemini-flash-latest 1M $1.50 $9.00
gpt5image openai/gpt-5.4-image-2 272k $8.00 $15.00

#🆓 Free tier (rate-limited)

Alias OpenRouter ID Ctx
kimi-free moonshotai/kimi-k2.6:free 256k
qwencoder-free qwen/qwen3-coder:free 1M
dsflash-free deepseek/deepseek-v4-flash:free 1M

#⚠️ Gotcha — Reasoning models and max_tokens

Reasoning models spend many tokens on internal "thinking" before responding. If max_tokens is too low, the model finishes with finish: length and NO visible answer (all tokens consumed by chain-of-thought).

Rule for skills:

  • General default: max_tokens: 2000
  • Reasoning models: max_tokens: 4000
  • Always allow user override

How to detect reasoning model (heuristic for skills):

  • Name contains thinking, reasoning, -r1, -o3, -o4
  • Family gpt-5.* (all 5.1+ variants are reasoning)
  • moonshotai/kimi-k2.6 (no suffix — reasoning by default)
  • moonshotai/kimi-k2-thinking (explicit)
  • qwen/qwen3-max-thinking (explicit)
  • deepseek/deepseek-r1*

Non-reasoning models (max_tokens 1000 is enough):

  • Family google/gemini-*-flash*
  • qwen/qwen3.6-flash, qwen/qwen3-coder-plus
  • deepseek/deepseek-v4-flash
  • mistralai/codestral-*
  • ibm-granite/granite-*
  • z-ai/glm-4.*-flash

#Updating prices / benchmarks

Prices on OpenRouter change. Verify via API before assuming:

curl -s https://openrouter.ai/api/v1/models | python3 -c "
import json, sys
data = json.load(sys.stdin)
for m in data['data']:
    if m['id'] == 'moonshotai/kimi-k2.6':
        print(m['pricing'])
"

To add a new model:

  1. Confirm ID via curl -s https://openrouter.ai/api/v1/models | python3 -c "import json,sys; [print(m['id']) for m in json.load(sys.stdin)['data']]"
  2. Define a short unique alias
  3. Add a row in the correct tier with ctx/price/benchmark
  4. Commit

#Refs