Skip to main content

LLM Models

Shisa LLM models use the same OpenAI-compatible chat completions endpoint — you select one with the model field in your request.

List models available to your key

Model availability is dynamic and filtered by API-key permissions. Always treat GET /openai/v1/models as the authoritative catalogue for the key you are using:

curl -s "https://api.shisa.ai/openai/v1/models" \
-H "Authorization: Bearer YOUR_API_KEY" | jq .

Choose a value from data[].id. When present, max_model_len reports that deployment's context limit.

Standard model tiers

The standard catalogue includes the following tiers. A model may be absent from your /models response if your key does not have access or the deployment changes.

Model IDTierParametersBest for
shisa-ai/shisa-v2.1-unphi4-14bFlash14BCost-efficient, high-throughput everyday tasks
shisa-ai/shisa-v2.1-llama3.3-70bPro (recommended)70BThe best balance of quality and speed for bilingual JA/EN work
shisa-ai/shisa-v2-llama3.1-405bFlagship405BMaximum quality; access by request

Shisa V2.1 Flash

shisa-ai/shisa-v2.1-unphi4-14b

The efficient tier. A 14B model built on a Phi-4 base, tuned for Japanese. Use it when latency and cost matter more than squeezing out the last few points of quality — chat widgets, classification, high-volume batch jobs.

Shisa V2.1 Pro

shisa-ai/shisa-v2.1-llama3.3-70b

The recommended default. A 70B model on a Llama 3.3 base. It posts a Japanese MT-Bench of 9.26 and a 73.1% Japanese average, making it the right choice for most production bilingual workloads. Read the Shisa V2.1 write-up for full evaluation details.

Shisa V2 405B

shisa-ai/shisa-v2-llama3.1-405b

The flagship. A 405B model on a Llama 3.1 base that reaches an 80.5% Japanese average — the highest quality Shisa offers.

Access by request

The 405B model is available to whitelisted accounts. Contact us to request access. Read the Shisa V2 405B write-up for details.

Context windows

Each tier has its own served context window. The standard deployments report approximately 16K tokens for Flash, 65K tokens for Pro, and 131K tokens for Shisa V2 405B. Read max_model_len from your authenticated /models response for the current limit.

Additional available models

We're partnering with Alibaba Cloud to expand Shisa's model roster with frontier coding and reasoning capabilities, including Qwen and GLM, optimized for the Japanese market based on our own benchmarks. This integration reflects our belief that the best developer experience combines open-source transparency with access to cutting-edge inference, all through a single API.

The following models are hosted in the Alibaba Cloud Singapore region.

Model NameModel IDDescription
GLM 5.2glm-5.2A flagship model for long-horizon tasks, reasoning, long-text comprehension, and code generation
Qwen 3.7 Flashqwen3.7-flashA multimodal model optimized for low latency and cost efficiency
Qwen 3.7 Maxqwen3.7-maxThe Qwen 3.7 flagship for complex agentic tasks
Qwen 3.7 Plusqwen3.7-plusA balanced multimodal model for coding, tool use, and productivity workflows

Next steps