Unverified50% confidenceBenchmarkExact time
在 Gemma、Granite、LFM 2.5 等不同模型和 tokenizer 上测试，启用约束解码后几乎没有额外性能开销
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/17/2026
First Seen
Valid until: 12/16/2026
Sources
Related Entities
Related Claims
UnverifiedGemma 4 31B 在评测中超越了智谱 GLM、DeepSeek V3.2 等同参数量级模型，并与 Gemini 5、Kimi 2.5 等大模型表现相当65% similarUnverifiedGemma 4 31B outperforms models of similar parameter counts including Zhipu GLM and DeepSeek V3.2 on benchmark scores.63% similarUnverifiedGemma 4 31B rivals much larger models like Gemini 5 and Kimi 2.5 on benchmark performance despite having far fewer parameters.61% similarUnverified如果 Gemma 3 12B 在结构化输出上频繁出错，建议换用 Qwen 2.5 系列或 Llama 3.1 8B/70B，因为它们对 Function Calling 有更好的原生支持60% similarUnverifiedGemma 3较小的模型规格（如4B、12B参数版本）可在消费级GPU上运行59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/928857API
curl https://kongchang.com/api/v1/knowledge/claims/928857MCP
get_claim(id=928857)