WizardLM 2 8x22B vs Claude Haiku 3 vs Llama-3.1-70B-Instruct

Claude Haiku 3 is the better all-round choice, scoring 46/100 against Llama-3.1-70B-Instruct (42) and WizardLM 2 8x22B (38). It leads on inputs & features and context window. The score weighs capability 50%, price 25%, inputs & features 15%, context window 10%. Capability uses the average of shared benchmarks (GPQA Diamond), because WizardLM 2 8x22B has no Capabilities Index score yet.

WizardLM 2 8x22B is cheaper at $0.493 input / $0.493 output per million tokens (median across 3 API providers). Claude Haiku 3 costs $0.25 input / $1.25 output per million tokens (median across 3 API providers); Llama-3.1-70B-Instruct costs $0.72 input / $0.72 output per million tokens (median across 8 API providers). At a typical mix of three input tokens to one output token, that is $0.493 per million tokens for WizardLM 2 8x22B versus $0.50 for Claude Haiku 3 (1× as much) and $0.72 for Llama-3.1-70B-Instruct (1.5× as much).

Not every model here has a Capabilities Index score, so we compare the average of shared benchmarks (GPQA Diamond): Llama-3.1-70B-Instruct 44.2%, WizardLM 2 8x22B 43.4% and Claude Haiku 3 36.3%. On individual benchmarks: GPQA Diamond — Llama-3.1-70B-Instruct 44.2%, WizardLM 2 8x22B 43.4%, Claude Haiku 3 36.3%.

There are no published SWE-bench Verified results for WizardLM 2 8x22B, Claude Haiku 3 and Llama-3.1-70B-Instruct yet, so there is no like-for-like coding score. On overall capability, Llama-3.1-70B-Instruct leads, which tends to carry over to coding, but test on your own codebase. Note that WizardLM 2 8x22B does not support tool calling, which most coding agents need.

Claude Haiku 3 has the largest context window at 200,000 tokens, against 128,000 for Llama-3.1-70B-Instruct and 65,536 for WizardLM 2 8x22B. Maximum output per response: WizardLM 2 8x22B up to 8,000, Claude Haiku 3 up to 4,096, Llama-3.1-70B-Instruct up to 4,096 tokens.

WizardLM 2 8x22B accepts text; Claude Haiku 3 accepts text, images and PDFs; Llama-3.1-70B-Instruct accepts text. Claude Haiku 3 handles the widest range of inputs.

WizardLM 2 8x22B and Llama-3.1-70B-Instruct publishes its weights (Apache-2.0) and can be self-hosted; Claude Haiku 3 is proprietary.

Llama-3.1-70B-Instruct is the newest, released Jul 23, 2024. WizardLM 2 8x22B came out Apr 15, 2024; Claude Haiku 3 came out Mar 13, 2024. Knowledge cutoff: Claude Haiku 3 Aug 31, 2023, Llama-3.1-70B-Instruct Dec 2023.

Each model gets a 0–100 score on capability (50%, independent benchmark results); price (25%, blended price per million tokens (3 input : 1 output), log scale); inputs & features (15%, image, PDF, audio and video input, tool calling, structured output and reasoning); context window (10%, maximum tokens per request, log scale). Dimensions missing for any model are dropped and the remaining weights rescaled, so every model is judged on the same evidence. Specs and prices come from public model listings and the labs’ own API pages; capability scores come from independent benchmark runs. Data updated Nov 2, 2026.