Z.ai GLM model, released December 22, 2025
GLM-4.7: Z.ai's coding model from the GLM-4.5 line
Released December 22, 2025 under the MIT license: a 355B-parameter (32B active) model with a 200K context, built as a coding partner. The specs, Z.ai's benchmarks, today's prices, and how it compares with GLM-4.6 and GLM-5.
GLM-4.7 is the last flagship built on the GLM-4.5 architecture (about 355B parameters, 32B active, 200K context). Z.ai pitched it as "your new coding partner": against GLM-4.6 it gains 5.8 points on SWE-bench Verified (73.8), 12.9 on SWE-bench Multilingual (66.7) and 16.5 on Terminal Bench 2.0 (41.0), and it scores 42.8 on Humanity's Last Exam with tools, up 12.4.
It also changed how the model thinks inside agents. Interleaved Thinking (thinking before every reply and tool call) was joined by Preserved Thinking, which keeps earlier reasoning blocks across turns in coding agents instead of re-deriving them, and Turn-level Thinking, which switches reasoning on or off per turn to trade accuracy for latency and cost. Z.ai also highlighted cleaner generated web pages and slides.
An independent reference, not affiliated with Z.ai. Checked against Z.ai's model card, docs and OpenRouter on October 4, 2026.
73.8
SWE-bench Verified (Z.ai)
355B
Total parameters, 32B active
200K
Context window (tokens)
MIT
License, open weights
Where GLM-4.7 stands today
GLM-5 replaced GLM-4.7 as Z.ai's flagship seven weeks later, with twice the parameters and higher scores on nearly every benchmark Z.ai compared. GLM-4.7 is still worth knowing for two reasons: at 355B it needs about half the hardware of a GLM-5 model to self-host (362 GB in FP8 against 756 GB), and its small sibling GLM-4.7-Flash is free on Z.ai's own API. For the best GLM results today, use GLM-5.3 or GLM-5.2.
Specifications
The GLM-4.7 spec sheet
From Z.ai's model card and the published config, with each row linked to its source.
What changed
GLM-4.7 against GLM-4.6
Agentic coding
SWE-bench Verified 68.0 to 73.8, SWE-bench Multilingual 53.8 to 66.7, Terminal Bench 2.0 24.5 to 41.0, LiveCodeBench v6 82.8 to 84.9.
Reasoning
Humanity's Last Exam 17.2 to 24.8 (30.4 to 42.8 with tools), HMMT February 2025 89.2 to 97.1, IMOAnswerBench 73.5 to 82.0.
Tool use and browsing
τ²-Bench 75.2 to 87.4, BrowseComp 45.1 to 52.0.
Thinking that survives a session
Preserved Thinking keeps reasoning across turns in coding agents, and Turn-level Thinking lets you switch it off for quick requests.
Better-looking output
Z.ai calls it vibe coding: cleaner, more modern web pages and slides with more accurate layout and sizing.
Benchmarks
GLM-4.7 benchmarks, as Z.ai published them
Every score and every comparison model comes from Z.ai's own model card. These are vendor-reported results: useful for direction, not independent proof.
| Benchmark | GLM-4.7 | GLM-4.6 | Kimi K2 Thinking | DeepSeek-V3.2 | Gemini 3.0 Pro | Claude Sonnet 4.5 | GPT-5-High | GPT-5.1-High |
|---|---|---|---|---|---|---|---|---|
| Reasoning | ||||||||
| MMLU-Pro | 84.3 | 83.2 | 84.6 | 85.0 | ||||
Where to run it
How to use GLM-4.7
Call it through Z.ai's API or a third-party host with any OpenAI-compatible SDK, use it inside a coding agent, or download the weights and serve them yourself.
- Z.ai API
Pricing
What GLM-4.7 costs on every host
On Z.ai's own API it costs $0.60 per 1M input tokens and $2.20 per 1M output tokens (cached input $0.11). 6 hosts list it on OpenRouter, and 3 of them charge less than Z.ai. Sorted cheapest first by blended price (three parts input to one part output).
| Host | Input /1M | Output /1M | Cached input /1M | Precision | Context | Max output |
|---|---|---|---|---|---|---|
DeepInfra | $0.40 | $1.75 | $0.08 | fp4 | 202,752 | 131,072 |
Venice | $0.4004 | $1.9292 | $0.0801 | fp4 | 198,000 | 16,384 |
Novita | $0.54 | $1.98 | $0.099 | fp8 | 204,800 | 131,072 |
Google | $0.60 | $2.20 | ||||
The GLM family
GLM-4.7 next to the other GLM models
Every GLM model with its own page here, newest first, with Z.ai's current price per 1M input / output tokens.
| Model | Released | Parameters | Context | License | Z.ai price (in / out) |
|---|---|---|---|---|---|
GLM-5.3 Flash Cheap, fast, 1M-context multimodal (the model behind Ox Alpha) | Aug 2026 | 320B total / 18B active | 1M | MIT | $0.15 / $0.50 |
GLM-5.3 Z.ai's current flagship: the strongest GLM for coding and agents, 1M context | Aug 2026 | 753B total (GLM-5.2 base) | 1M | GLM-5.3 License | $1.40 / $4.40 |
GLM-5.2 The newest GLM flagship under plain MIT: 1M context, long-horizon coding | Jun 2026 | 744B / 40B active | 1M | MIT | $1.40 / $4.40 |
FAQ
GLM-4.7, answered
What is GLM-4.7?
Is GLM-4.7 open source?
How much does GLM-4.7 cost?
What is the difference between GLM-4.7 and GLM-4.6?
What is the difference between GLM-4.7 and GLM-5?
What is GLM-4.7-Flash?
Build something with GLM-4.7
You have seen the specs, the scores and the price. Turning a model into a product people use is the harder part, and it is the part you can skip wiring up yourself.