GLM 5.2 vs GLM 5.3 Flash
Close to a tie overall. GLM 5.3 Flash costs 80% less for the same work, and both read a 1M-token context.
GLM 5.2
by Z.ai · #15 of 67 overall
- Context
- 1M
- Input / 1M
- $0.75
- Output / 1M
- $2.40
GLM 5.3 Flash
by Z.ai · #13 of 67 overall
- Context
- 1M
- Input / 1M
- $0.15
- Output / 1M
- $0.50
Better overall
About even
Grade 82.1 vs 82.9
Cheaper
GLM 5.3 Flash
80% less, $0.24 vs $1.16 blended
Larger context
Same
1M vs 1M tokens
Terminal-Bench 2.1
GLM 5.3 Flash
81% vs 84.3%
Head to head
Public benchmarks. Longer is better; each value links to its source.
GLM 5.2GLM 5.3 Flashv = reported by the model’s maker. — = no public score yet. Arena bars span the range across every model we grade.
Speed through FullsendAI
Measured through FullsendAI, from one location, Oct 15, 2026, n=20 (5 for long runs).
GLM 5.2GLM 5.3 Flash
Time to first tokenmedian, tick at p90 · lower is faster
Output speed1,500-token replies, tick at p90
First-try success — for GLM 5.2 and 47/55 for GLM 5.3 Flash. FullsendAI retries busy responses automatically, so a missed first try costs a moment, not an error.
How we measured
Streamed chat completions with default settings, the two models alternating run by run. Time to first token runs from sending the request to the first streamed token, connection included, over short prompts. Output speed is output tokens a second after the first token, on long replies. One location and one afternoon, so treat small differences as noise.
What it costs you
Drag to match your monthly usage.
GLM 5.3 Flash saves you $196 a month at this usage, $2,352 a year.
Capabilities
| Capability | GLM 5.2 | GLM 5.3 Flash |
|---|---|---|
| Reasoning levels through FullsendAI | No | Yes |
| Reads images through FullsendAI | No | Yes |
| Open weights | — | Yes |
| Tool calling | — | Yes |
| JSON output | — | Yes |
| Tool calls that worked (measured) | — | 10/10 |
| Valid JSON replies (measured) | — | 10/10 |
Our coding quick check (ten small Python functions against hidden tests; not a benchmark): — for GLM 5.2, 10/10 for GLM 5.3 Flash.
Maker-published benchmarks
As each maker reports them, with their own harnesses and settings: not graded, and not strictly comparable across makers.
Each score links to where its maker published it. “chart” marks a number read from a chart image.
At a glance
| Attribute | GLM 5.2 | GLM 5.3 Flash |
|---|---|---|
| Made by | Z.ai | Z.ai |
| Availability | Coming soon | Coming soon |
| Overall grade | Excellent · 82.1 | Excellent · 82.9 |
| Overall rank | #15 of 67 | #13 of 67 |
| Input price | $0.75 per 1M | $0.15 per 1M |
| Output price | $2.40 per 1M | $0.50 per 1M |
| Context window | 1,048,576 tokens | 1,048,576 tokens |
| Reasoning levels | No | Yes |
| Reads images | No | Yes |
| Model ID | glm-5.2 | glm-5.3-flash |
Model card: what the makers publish
| Fact | GLM 5.2 | GLM 5.3 Flash |
|---|---|---|
| Released | — | Aug 26, 2026 |
| Made by | — | Z.ai (zai-org; Zhipu AI) |
| Licence | — | MIT License |
| Commercial use | — | Allowed (MIT License) |
| Open weights | — | Yes |
| Parameters | — | 320B |
| Active parameters | — | 18B |
| Mixture of experts | — | Yes: 288 routed experts + 1 shared, 8 experts per token, first 3 layers dense |
| Architecture | — | Hybrid sparse + linear attention (linear attention for local dependencies, sparse attention with a lightweight indexer for global context; IndexPool compresses 4 indexer keys into 1), Manifold-Constrained Hyper-Connections (mHC), MTP layer, ViT vision encoder; 45 layers. Z.ai: attention compute and KV cache reduced 3.01x and 4.44x vs GLM-5.3. |
| Context window | — | 1,048,576 tokens |
| Max output | — | 128K (Z.ai API) |
| Input | — | Text, Image, Video, File |
| Output | — | Text |
| Reasoning | — | Thinking with reasoning_effort low/high/max (default max if omitted). On Z.ai's API thinking is forced (thinking.type only supports 'enabled'; cannot be disabled). Chat template clear_thinking defaults to false; pass clear_thinking=true for chat. Interleaved and preserved thinking supported. |
| Tool calling | — | Yes |
| JSON output | — | Structured output / JSON mode via response_format {'type':'json_object'} |
| Recommended sampling | — | temperature 1, top_p 0.95, reasoning_effort max; thinking.clear_thinking false; stream + tool_stream for streaming |
| Training data | — | Newly trained base model; 30T-token multimodal pre-training corpus |
| Other features | — | Context caching; visual coding loop (screenshots/recordings to apps), computer use; GLM-5.3-FlashX variant at ~200 tokens/s on Z.ai; MTP layer (speculative decoding) in architecture |
| Runs locally with | — | SGLang, vLLM, TokenSpeed, Transformers, KTransformers, Unsloth |
Sources
- Released: GLM 5.3 Flash
- Made by: GLM 5.3 Flash
- Licence: GLM 5.3 Flash
- Commercial use: GLM 5.3 Flash
- Open weights: GLM 5.3 Flash
- Parameters: GLM 5.3 Flash
- Active parameters: GLM 5.3 Flash
- Mixture of experts: GLM 5.3 Flash
- Architecture: GLM 5.3 Flash
- Context window: GLM 5.3 Flash
- Max output: GLM 5.3 Flash
- Input: GLM 5.3 Flash
- Output: GLM 5.3 Flash
- Reasoning: GLM 5.3 Flash
- Tool calling: GLM 5.3 Flash
- JSON output: GLM 5.3 Flash
- Recommended sampling: GLM 5.3 Flash
- Training data: GLM 5.3 Flash
- Other features: GLM 5.3 Flash
- Runs locally with: GLM 5.3 Flash
Try both with one key
Same endpoint, same key, both hosted in the USA. Change one word to switch models, in Claude Code, Cursor, Cline or your own code.
Get your API key# Ask either model the same thing curl https://fullsendai.com/v1/chat/completions \ -H "Authorization: Bearer $FULLSEND_API_KEY" \ -d '{"model": "glm-5.2", "messages": [...]}' # …then swap one word -d '{"model": "glm-5.3-flash", "messages": [...]}'
Questions
Is GLM 5.2 or GLM 5.3 Flash better?
They grade within two points of each other (82.1 and 82.9), so pick on price and the benchmarks that match your work.
Which is cheaper, GLM 5.2 or GLM 5.3 Flash?
GLM 5.3 Flash: $0.15 in and $0.50 out per million tokens, against $0.75 and $2.40 through FullsendAI.
Which has the larger context window?
Neither: both take 1,048,576 tokens.
Can I use both with one API key?
Yes, once they launch: both are coming soon, at the prices shown, and both will use the same FullsendAI key.
Compare with others
Scores: LMArena (CC BY 4.0), Scale AI SWE-Bench Pro, Terminal-Bench, and model makers’ published results where marked v. Prices are FullsendAI’s, per million tokens. Data as of Oct 28, 2026.