Live pricing on OneInfer
Pulled from the OneInfer catalog (OpenRouter and direct provider routes) at request time — rates can change. Capture the timestamp with any benchmark or production decision.
| Model | Provider | Input $/1M tokens | Output $/1M tokens |
|---|---|---|---|
| GLM-5.3 | zai | $140.00 | $440.00 |
| GLM-5.3 | zai | $15.00 | $50.00 |
| GLM-5.3 | zai | $37.00 | $125.00 |
| GLM-5.3 | akashml | $117.00 | $396.00 |
| GLM-5.3 | novita | $15.00 | $50.00 |
| GLM-5.3 | together_ai | $15.00 | $50.00 |
| GPT-5.6 Sol | openai | $125.00 | $1000.00 |
| GPT-5.6 Sol | openai | $400.00 | $2000.00 |
Decision snapshot
| Decision factor | GLM-5.3 | GPT-5.6 Sol |
|---|---|---|
| Independent intelligence score | 60 | 61 |
| Current input price / 1M tokens | $1.40 | $4.00 promotional |
| Current cached input / 1M tokens | $0.26 | $0.40 |
| Current output price / 1M tokens | $4.40 | $20.00 promotional |
| Context window | 1M tokens | 1.05M tokens |
| Maximum output | 128K tokens | 128K tokens |
| Input modalities | Text | Text and images |
| Model access | API; weights announced for release | Managed API; proprietary weights |
How we keep this comparison honest
- Match the tested modes: the independent table compares GLM-5.3 max with GPT-5.6 Sol max on Artificial Analysis Intelligence Index v4.1.1.
- Separate measured cost from today's list price: Artificial Analysis recorded Sol at its earlier $5 input and $30 output rates; OpenAI now lists promotional $4 and $20 rates.
- Keep independent and vendor evidence separate: Artificial Analysis supports the main verdict, while Z.ai coding and security scores appear in their own vendor-reported table.
- Count token efficiency as its own result: Sol used 70M evaluation output tokens versus GLM-5.3's 170M, even though GLM-5.3 had the lower measured task cost.
- Apply the long-context price rule: OpenAI charges 2x input and 1.5x output rates to the full Sol request when input exceeds 272K tokens.
Independent performance and efficiency
Artificial Analysis ran both models at max reasoning effort. Its recorded task costs reflect provider prices at evaluation time; use the current first-party prices above when estimating a new production workload.
| Artificial Analysis metric | GLM-5.3 (max) | GPT-5.6 Sol (max) |
|---|---|---|
| Intelligence Index v4.1.1 | 60 | 61 |
| Cost per Intelligence Index task | $0.68 | $1.23 |
| Output speed | 84.1 tokens/s | 74.4 tokens/s |
| Total evaluation output tokens | 170M | 70M |
What the independent results mean
- GPT-5.6 Sol leads the composite intelligence score by one point, so this is not evidence of a universal quality gap.
- GLM-5.3 cost about 45% less per evaluated task and generated output about 13% faster in this test.
- GPT-5.6 Sol used about 59% fewer output tokens, which can reduce latency, review time, and downstream context growth.
- Run both on a private prompt set because effort level and workload mix can change the quality-cost winner.
Vendor-reported coding and security benchmarks
Z.ai reports the following results for the named benchmark rows. These provide useful directional evidence but are vendor-reported, not independently verified results.
| Benchmark | GLM-5.3 | GPT-5.6 Sol |
|---|---|---|
| Terminal-Bench 2.1 | 88.2 | 88.8 |
| Terminal-Bench 3.0 | 28.3 | 34.6 |
| DeepSWE v1.1 | 66.9 | 72.7 |
| CyberGym | 84.5 | 83.6 |
| ExploitBench | 54.4 | 76.5 |
Ready to test the workflow?
Create account & add creditsCapabilities and API behavior
| Capability | GLM-5.3 | GPT-5.6 Sol |
|---|---|---|
| Reasoning effort | Low, high, and max; always enabled | None, low, medium, high, xhigh, and max; medium default |
| Function calling and structured output | Supported | Supported |
| Vision input | Not supported by the base GLM-5.3 model | Supported |
| Hosted tools | Provider-dependent | Web search, file search, code interpreter, hosted shell, computer use, MCP, and more |
| Fine-tuning | Verify current provider support | Not supported |
Long-context pricing changes the calculation
Both models support roughly one million tokens of context, but their billing rules differ. OpenAI documents promotional Sol rates of $4 input and $20 output per million tokens through at least November 21, 2026. Once a request exceeds 272K input tokens, the entire Sol request is billed at $8 input and $30 output per million tokens. Recalculate long-document and repository-agent costs at the expected prompt size rather than using headline rates.
Deployment and model-control tradeoffs
GPT-5.6 Sol is proprietary and served through managed APIs. Z.ai announced that GLM-5.3 weights would be released on August 28, 2026; at this page's August 27 verification point they were not yet published. Verify the released weights, license, runtime support, and hardware requirements before treating GLM-5.3 as a self-hosting option.
Which model should you choose?
| Workload or priority | Recommended starting point | Why |
|---|---|---|
| Highest measured general capability | GPT-5.6 Sol | It holds a narrow independent index lead. |
| Budget-sensitive text agents | GLM-5.3 | Its standard first-party token prices and measured task cost are lower. |
| Concise agent output | GPT-5.6 Sol | It used substantially fewer output tokens in the independent evaluation. |
| Image or screenshot understanding | GPT-5.6 Sol | Sol accepts image input; base GLM-5.3 is text-only. |
| Very long prompts above 272K tokens | Benchmark total cost | Sol applies higher rates to the entire long-context request. |
| Future self-hosting or model control | GLM-5.3 | Z.ai announced an open-weight release; confirm availability and license before deployment. |
| Deep software or security tasks | Test both | Sol leads several reported coding and exploitation rows; GLM narrowly leads CyberGym. |
Frequently asked questions
Can I try GLM-5.3 before integrating it?
Use the OneInfer GLM-5.3 launcher to open a prepared prompt in the authenticated playground. Availability is checked against the current model catalog.
How should I treat benchmark claims?
Check the provenance label and harness version. Vendor-reported and independently verified results are deliberately shown as different evidence classes.
Is GLM-5.3 cheaper than GPT-5.6 Sol?
At first-party prices verified on August 27, 2026, GLM-5.3 has lower standard input and output token prices. Actual task cost depends on token use, caching, reasoning effort, tools, and whether a Sol prompt exceeds the 272K long-context pricing threshold.
Which model is more token-efficient?
In Artificial Analysis Intelligence Index v4.1.1 at max effort, GPT-5.6 Sol produced 70M output tokens versus GLM-5.3's 170M while scoring 61 versus 60. Treat that as evaluation-specific evidence and reproduce it on your workload.
Can both models process images?
No. GPT-5.6 Sol supports text and image input. The base GLM-5.3 model accepts text only.
Do both models support approximately one million tokens of context?
Yes. GLM-5.3 documents 1M tokens and GPT-5.6 Sol documents 1.05M. Sol requests above 272K input tokens use higher rates for the full request.
Put GLM-5.3 to work
Fund a controlled evaluation, start with a prepared prompt, and measure quality and cost on your own workload.