GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
armand0e
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
armand0e
Model Tree
Base
Qwen/Qwen3.5-2B
Adapter
this model
Input Modalities
Text
Image
Video
Output Modalities
Text
Supported Functionality
Dedicated Endpoints
complete_skippedTrueoutputs/qwen-capability-light/stage2-capability-step-sft/checkpoint-260outputs/qwen-capability-light/stage2-capability-step-sft/checkpoint-2602.5278091430664062outputs/qwen-capability-light/stage2-capability-step-sft/trainer_state.json2601.52169382572174071.6552110910415650.64226382970809942.09362220764160162602.62878489494323732.52780914306640622.52780914306640622.6287848949432373142.52780914306640622603794.8465Generated files:
training_config.jsonstage_report.jsonloss_history.csvloss_curve.svgeval_loss_history.csveval_loss_curve.svg
armand0e/qwen3.5-capability-light-behavior-seed-lorastage3-capability-dpoQwen/Qwen3.5-2Bdata/assembled/sft_qwen_next_actions_capability_light.jsonldata/eval/eval_next_actions_with_retention.jsonl16 / 16 / 0.08e-07armand0e/qwen3.7-max-pi-tracesarmand0e/badlogicgames-pi-mono-opus-filteredarmand0e/gpt-5.5-agentarmand0e/gpt-5.5-chatTeichAI/claude-4.5-opus-high-reasoning-250xTeichAI/Claude-Opus-4.6-Reasoning-887xjson
{"messages": [{"content": "User/task context:\nuser: Give only the answer: (98 + 459) - 34 = ?","role": "user"},{"content": "523","reasoning_content": "Expression evaluates to 523. The answer should return only the numeric answer, while making sure no extra words.","role": "assistant"}],"metadata": {"expected": 523,"family": "arithmetic","key": "gap/arithmetic/00000","source": "gap_capability_pack"},"source": "gap_capability_pack"}
The exact stage command and package versions are in training_config.json.
1.0armand0e/qwen3.5-capability-light-capability-step-merged-16bit