Гэрэг · Qwen 3.8 capability showcase

Өөрөө өөрийгөө шалгадаг мультимодаль pipeline. Энд харуулсан бүх төлөв амьд probe-н үр дүн.

Монгол TTS: qwen-audio-3.0-tts-plus ships zh/en voices only; Mongolian is rendered phonetically (e.g. 'Гэрэг' -> 'Грек'). Treat Mongolian TTS output as approximate and measure it.

API

Model registry

token-planmodelrolepayg
✅qwen3.8-maxflagship reasoning + native vision, 1M ctx✅
✅qwen3.8-max-previewflagship preview channel⛔
✅qwen3.8-flashfast/cheap tier, 1M ctx, vision✅
✅qwen3.7-maxprevious-gen flagship✅
✅qwen3.7-plusprevious-gen balanced, vision✅
✅qwen3.6-flashprevious-gen fast tier✅
✅glm-5.2third-party, available on both plans✅
⛔glm-5.3third-party, payg only✅
✅deepseek-v4-prothird-party, available on both plans✅
✅deepseek-v4-flash-0731third-party fast tier✅
✅deepseek-v4-pro-0813third-party snapshot✅
⛔deepseek-v4.1-flashthird-party, payg only✅
⛔qwen3.8-omni-flashomni multimodal, 1M ctx✅
⛔qwen3-vl-plusdedicated vision-language model✅
⛔qwen3-coder-pluscode model✅
⛔qwen3-coder-nextcode model, next channel✅
⛔qwen3.5-plusolder-gen balanced✅
🟡qwen3.6-plusprevious-gen balanced✅
⛔qwen3-max-2026-01-23dated snapshot✅
🟡deepseek-v3.2third-party✅
🟡glm-5.1third-party✅
🟡kimi-k2.7-codethird-party code model✅
🟡kimi-k2.6third-party⛔
🟡kimi-k2.5third-party⛔
🟡glm-5third-party⛔
🟡MiniMax-M2.5third-party⛔
⛔glm-4.7third-party⛔
⛔qwen-deep-researchlong-horizon research agent⛔
✅wan2.7-imageimage generation + instruction editing (Token Plan)—
⛔qwen-image-3.0newer image model✅
⛔qwen-image-3.0-pronewer image model, pro tier✅
⛔qwen3.7-text-embedding1024-dim embeddings; enables vector RAG✅
✅happyhorse-1.1-t2vtext-to-video, async task API✅
✅qwen-audio-3.0-tts-plusspeech synthesis (zh/en voices only)—
✅qwen-audio-3.0-asr-flashspeech recognition from a public URL—
⛔qwen-audio-3.1-tts-flashlow-latency TTS⛔