Huawei's 92B MoE lightweight (6B active) for fast agents and 512K context work, Ascend-native with slow/fast thinkingopenPangu 2.0 Flash by huawei is currently listed from a single provider. No official API price is listed for it; the prices below come from third-party channels.
The context window is 524,288 tokens, with an output limit of 131,072 tokens. It supports reasoning and tool use. The weights are open, so it can also be self-hosted or served through a gateway of your choice. Its training knowledge cuts off at 2026-06.
| Provider | Tier | Input | Output | Cache read | Cache write | Context | Output limit | Status |
|---|---|---|---|---|---|---|---|---|
| Huawei Cloud MAAS | Gateway | $0.12 | $0.24 | $0.03 | — | 524,288 | 131,072 |
Sorted by blended price (input×0.75 + output×0.25) asc. The official channel always shows regardless of rank. Whether a gateway's low price is actually usable can't be verified.
Interleaved thinking (reasoning between tool calls) is declared by 1 of 1 providers.
Price history accumulates from each data sync; currently only 1 sample(s) (2026-11-15). Each future sync adds a point, and once accumulated a line is drawn here.
| Name | Conditions | Score | Metric | Source |
|---|---|---|---|---|
| AIME 2026 | variant: thinking | 93.3 | Avg@16 | Source ↗ |
| SWE-bench Verified | variant: thinking | 63.1 | Avg@3 | Source ↗ |
| LiveCodeBench V6 | variant: thinking | 85.1 | Avg@3 | Source ↗ |
| GPQA-Diamond | variant: thinking | 83.7 | Avg@4 | Source ↗ |
The same benchmark scores very differently across harness / dataset, so the qualifying conditions must be shown together.