Questions? hello@viberation.devGet supportBlogDocsChangelog
Get started

Z.ai's open-weight family, behind the cheap coding subscriptions.

GLM is the model family from Z.ai, and the reason its coding plan turns up in every "cheaper than Claude Code" thread — the weights are published, plenty of providers serve them, and the Flash models are some of the cheapest usable coding models anywhere. The 5.3 line runs to a million tokens of context, with Flash variants that also read images and video. Z.ai's own chat app is listed separately under Chats.

At a glance — GLM 5.3 Flash

Cost
$0.045 in · $0.14 out per 1M tokensBudget
Memory
1.3M tokens — about 1,950 pages
Reads
Text · Images · Video
Can do
  • Use tools (agents)
  • Structured output
  • Thinking mode
Good at
Coding: better than 80% · Agentic work: better than 91% · Overall: better than 79%
Reliability
99.90% uptime, last 24h
Released
Aug 26, 2026
Advanced details
API model ID
z-ai/glm-5.3-flash
Max output
128K tokens
Cached input
$0.01 per 1M
Tokenizer
Other
Coding index
71.5
Agentic work index
50.9
Overall index
41.8

Providers

The same model, hosted by different companies. OpenRouter picks one per request and falls back to the next if it fails.

ProviderInput /1MOutput /1MCached /1MMax outputUptime 24h
InferenceNet · fp4$0.045$0.14$0.01128K99.41%
Sail Research · us$0.045$0.60$0.029131K99.64%
Sail Research · fp8$0.045$0.60$0.029131K99.85%
OpenInference · fp4$0.05$0.50$0.015944K99.90%
Relace$0.07$0.28$0.02131K99.85%
DeepInfra · fp4$0.075$0.25$0.015131K99.77%
Wafer$0.089$0.35$0.03944K99.77%
GMICloud · fp8$0.09$0.30$0.018944K99.38%
Morph$0.098$0.343$0.02944K96.73%
Inceptron · fp8$0.11$0.45$0.07944K99.28%
Decart · fp4$0.128$0.425$0.026944K98.60%
Phala · fp8Degraded$0.128$0.425$0.026131K98.31%
Novita · fp8$0.132$0.44$0.026131K99.73%
StreamLake · fp8$0.141$0.47$0.028128K99.46%
Modal · nvfp4$0.15$0.50$0.03944K99.74%
BaseTen · fp8$0.15$0.50$0.03131K99.16%
Near AI · fp8$0.15$0.50$0.035131K99.52%
Crusoe · fp4$0.15$0.50$0.03944K99.42%
CoreWeave · nvfp4$0.15$0.50$0.05944K98.51%
AtlasCloud · fp8$0.15$0.50$0.03131K99.56%
Fireworks$0.15$0.50$0.03944K99.53%
Friendli$0.15$0.50$0.03944K99.55%
SiliconFlow · fp8$0.15$0.50$0.03262K99.75%
DigitalOcean$0.15$0.50$0.03944K99.14%
Together$0.15$0.50$0.03944K99.87%
Reka$0.15$0.50$0.03236K99.49%
Parasail · fp8$0.15$0.50$0.03944K99.30%
BaseTen · fp8$0.15$0.50$0.03131K99.24%
Venice$0.15$0.50$0.03131K99.29%
Io Net · fp8$0.15$0.50$0.0366K99.06%
Z.AI · fp8$0.15$0.50$0.03131K99.07%
NextBit · fp8$0.165$0.55$0.033128K99.83%
Cloudflare$0.30$1.00$0.031.2M99.86%

Supported parameters

  • frequency_penalty
  • include_reasoning
  • logit_bias
  • logprobs
  • max_tokens
  • min_p
  • parallel_tool_calls
  • presence_penalty
  • reasoning
  • reasoning_effort
  • repetition_penalty
  • response_format
  • seed
  • stop
  • structured_outputs
  • temperature
  • tool_choice
  • tools
  • top_k
  • top_logprobs
  • top_p

Live from OpenRouter, refreshed hourly. Scores from Artificial Analysis.

Key info

Pricing
Open source
Category
Models
Best for
Intermediate
Made by
Z.ai
Weights
Open
Good for
Coding on a budget

More in Models

  • France's frontier lab, with the small models everyone self-hosts.

    Models
  • Meta's current line, built for long multi-agent runs.

    Models
  • The open-weight family everyone else gets benchmarked against.

    Models
  • Google's open models — Gemini's research, small enough to self-host.

    Models