Released: 2024-04-15

WizardLM-2 8x22B

Microsoft AI Mixture-of-Experts model trained with multi-agent reinforcement learning for complex multi-turn dialogue.

Intelligence Index
82.0/ 100
Calibrated multi-domain composite
Throughput Speed
68tok/sec
Output streaming throughput
Time To First Token
380ms
Initial chunk server latency
LiveBench Contam-Free
71.5%
September 2026 suite score

Sourced Benchmark Evaluations (61)

Verified performance across authoritative benchmarks with provenance tracking.
BenchmarkStatusScoreTrustDateSource TypeProvenance
LiveBench (Continuous)Active71.5%
94
2025-02-01independentSource ↗
ARC-AGI-3Active52.7%
93
2024-05-20independentSource ↗
ARC-AGI-2Nearing Saturation63.2%
75
2024-05-13independentSource ↗
FrontierMathActive22.4%
98
2024-05-10independentSource ↗
PutnamBenchActive41.6%
90
2024-05-09independentSource ↗
Omni-MATHDeprecated59.2%
10
2024-05-07independentSource ↗
SWE-LancerDeprecated51%
20
2024-05-07independentSource ↗
SoohakActive60.8%
98
2024-05-05independentSource ↗
FrontierCodeActive43.5%
98
2024-05-05independentSource ↗
Humanity's Last ExamActive35.6%
93
2024-05-03independentSource ↗
StructFlowBenchNearing Saturation64.1%
75
2024-05-03independentSource ↗
AIMENearing Saturation76%
75
2024-05-03independentSource ↗
OlympiadBenchActive65.6%
80
2024-05-02independentSource ↗
FireBenchActive63.4%
95
2024-05-01independentSource ↗
Terminal-BenchActive52.5%
80
2024-05-01independentSource ↗
MathArenaActive64%
99
2024-05-01independentSource ↗
OpenAI-MRCRActive68.9%
80
2024-04-30independentSource ↗
CodeContestsDeprecated48.8%
10
2024-04-30independentSource ↗
SimpleQAActive42.6%
90
2024-04-30independentSource ↗
ComplexBenchNearing Saturation61.9%
80
2024-04-29independentSource ↗
SWE-bench ProActive46.5%
90
2024-04-29independentSource ↗
LongBench-ProActive67.2%
80
2024-04-29independentSource ↗
MultiPL-EDeprecated69%
10
2024-04-28independentSource ↗
ZeroSCROLLSNearing Saturation65.6%
78
2024-04-28independentSource ↗
AgentIFNearing Saturation51.3%
80
2024-04-27independentSource ↗
MATHSaturated78.4%
25
2024-04-27independentSource ↗
LiveCodeBenchSaturated61.5%
39
2024-04-27independentSource ↗
InfiniteBenchNearing Saturation69.7%
65
2024-04-27independentSource ↗
MMLU-ProNearing Saturation75.4%
75
2024-04-26independentSource ↗
MercuryDeprecated66%
10
2024-04-26independentSource ↗
SCROLLSNearing Saturation65.6%
65
2024-04-26independentSource ↗
RULERNearing Saturation75.4%
84
2024-04-25independentSource ↗
TabMWPSaturated73.6%
25
2024-04-25independentSource ↗
CyberSecEvalDeprecated67.5%
10
2024-04-25independentSource ↗
MMLU-ReduxSaturated80.4%
25
2024-04-24independentSource ↗
QuALITYNearing Saturation72.2%
78
2024-04-24independentSource ↗
SWE-bench VerifiedSaturated58.5%
25
2024-04-24independentSource ↗
MGSMNearing Saturation76%
65
2024-04-23independentSource ↗
LongBenchNearing Saturation70.5%
65
2024-04-23independentSource ↗
SWE-benchDeprecated51%
10
2024-04-23independentSource ↗
AGIEvalSaturated69.7%
25
2024-04-22independentSource ↗
QASPERNearing Saturation68.1%
65
2024-04-22independentSource ↗
TruthfulQASaturated72.2%
25
2024-04-22independentSource ↗
SuperGLUESaturated82.6%
25
2024-04-21independentSource ↗
GSM-HardDeprecated76.8%
10
2024-04-21independentSource ↗
MT-BenchNearing Saturation7.8%
75
2024-04-20independentSource ↗
BIG-Bench HardSaturated79.4%
25
2024-04-20independentSource ↗
CodeXGLUEDeprecated70.5%
23
2024-04-20independentSource ↗
MBPP+Nearing Saturation76.5%
65
2024-04-19independentSource ↗
WinoGrandeSaturated84.2%
28
2024-04-19independentSource ↗
IFEvalDeprecated73.9%
25
2024-04-19independentSource ↗
GPQANearing Saturation72.2%
78
2024-04-19independentSource ↗
HumanEval+Saturated81%
25
2024-04-18independentSource ↗
Needle-in-a-HaystackSaturated96.4%
44
2024-04-18independentSource ↗
CommonsenseQASaturated85.1%
28
2024-04-18independentSource ↗
GSM8KSaturated86.4%
25
2024-04-17independentSource ↗
GLUESaturated87.5%
28
2024-04-17independentSource ↗
MBPPSaturated79.5%
25
2024-04-17independentSource ↗
HumanEvalSaturated85.5%
25
2024-04-16independentSource ↗
MMLUSaturated83.6%
25
2024-04-16independentSource ↗
ARC-AGISaturated76.1%
25
2024-04-16independentSource ↗
Local Hardware Execution (Ollama)Hardware Compatible
# 1. Pull and execute model locally with Ollama
ollama run wizardlm-2-8x22b

# 2. Or invoke via local OpenAI-compatible endpoint
curl http://localhost:11434/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "wizardlm-2-8x22b",
    "messages": [{"role": "user", "content": "Analyze reasoning chains on AIME 2026"}]
  }'

Evaluation & Sourcing Notes

Scores listed for WizardLM-2 8x22B represent verified evaluations extracted from official research papers, independent evaluation suites (HELM, LMSYS, OpenCompass, LiveBench), and verified audit reports.

All benchmarks marked as saturated or deprecated reflect historical performance where the benchmark no longer provides active discriminative power.