fal-ai/wizardlm-2-8x22b

fal-ai/wizardlm-2-8x22b is a efficient-tier language model from fal.ai with 141B parameters. fal-ai/wizardlm-2-8x22b is an open-weights model, meaning the weights are publicly available for self-hosting, fine-tuning, and deployment on your own infrastructure — not just through fal.ai's API. The WizardLM family is fal.ai's model series for this capability tier.

fal-ai/wizardlm-2-8x22b is priced at $0.5/1M input tokens and $0.5/1M output tokens. Output tokens are 1× more expensive than input tokens — this ratio reflects the sequential nature of autoregressive generation, which requires a separate forward pass per output token versus parallel processing of all input tokens simultaneously. At a typical usage pattern of 1,000 input tokens and 500 output tokens per call, fal-ai/wizardlm-2-8x22b costs approximately $0.75 per 1,000 API calls. fal.ai does not publish a prompt caching rate for this model, so repeated long contexts are billed at the standard $0.5/1M input rate.

Its 66K-token context window is sufficient for most production use cases, handling documents up to ~25,000 words or multi-turn conversations spanning dozens of exchanges. Median latency (P50 time-to-first-token) is approximately 500ms — suitable for interactive applications where sub-second response is acceptable. Throughput is approximately 120 tokens/second — adequate for streaming at comfortable reading speed. Maximum output length is 8K tokens — adequate for most generation tasks including essays, code functions, and detailed responses.

fal-ai/wizardlm-2-8x22b is best suited to chat, coding. Supported capabilities include streaming token output for real-time UIs. When evaluating this model for a specific workload, the most important factors are: context window (does your input fit?), output price (how long are your generations?), and latency (is this interactive or batch?). The pricing table above shows how fal-ai/wizardlm-2-8x22b compares to similar models across these dimensions.

fal-ai/wizardlm-2-8x22b sits in the efficient tier — optimized for speed and cost while maintaining solid quality, ideal for high-volume or latency-sensitive workloads. As an open-weights model, fal-ai/wizardlm-2-8x22b can be self-hosted on GPU compute (see the GPU rental comparison on this site) to eliminate per-token API costs entirely — the break-even point versus API pricing depends on your volume and the GPU hourly rate. When comparing fal-ai/wizardlm-2-8x22b to alternatives, use the similar models section on this page to find models at comparable price points with different capability tradeoffs. The intelligence score (55/100) reflects aggregate benchmark performance — a useful proxy for general capability, though real-world performance varies significantly by task type. Always test with your specific prompts and use case before committing to a model for production.