IFBench

General

Source ↗ Higher is better

Rank	Model	Provider	Score
1	Grok 4.3 (medium)	SpaceXAI	83.3%
2	Grok 4.20 0309 (Reasoning)	SpaceXAI	82.9%
3	MiniMax: MiniMax M3	MiniMax	82.9%
4	Nemotron 3 Ultra 550B A55B (Reasoning)	NVIDIA	81.4%
5	Grok 4.3 (high)	SpaceXAI	81.3%
6	Grok 4.20 0309 v2 (Reasoning)	SpaceXAI	81.2%
7	Grok 4.3 (low)	SpaceXAI	81.0%
8	Qwen: Qwen3.7 Max	Alibaba	80.5%
9	Nemotron Cascade 2 30B A3B	NVIDIA	80.4%
10	MiMo-V2.5-Pro (Reasoning)	Xiaomi	79.9%
11	Nova 2.0 Pro Preview (low)	Amazon	79.6%
12	DeepSeek V4 Flash 0420 (Reasoning, Max Effort)	DeepSeek	79.2%
13	Nova 2.0 Pro Preview (medium)	Amazon	79.0%
14	Qwen3.5 397B A17B (Reasoning)	Alibaba	78.8%
15	Gemini 3 Flash Preview (Reasoning)	Google	78.0%
16	Qwen: Qwen3.7 Plus	Alibaba	78.0%
17	GPT-5.2 Codex (xhigh)	OpenAI	77.6%
18	Google: Gemini 3.1 Flash Lite	Google	77.2%
19	Google: Gemini 3.1 Pro Preview	Google	77.1%
20	Qwen: Qwen3.6 Max PreviewHistorical	Alibaba	76.6%
21	DeepSeek V4 Pro 0424 (Reasoning, Max Effort)	DeepSeek	76.5%
22	Gemini 3.5 Flash (high)	Google	76.3%
23	GLM-5.1 (Reasoning)	Z.ai	76.3%
24	Kimi K2.6 (Reasoning)	Moonshot AI	76.0%
25	GPT-5.4 nano (xhigh)	OpenAI	75.9%
26	Muse Spark	Meta	75.9%
27	GPT-5.5 (xhigh)	OpenAI	75.9%
28	Qwen3.5 122B A10B (Reasoning)	Alibaba	75.7%
29	MiniMax: MiniMax M2.7	MiniMax	75.7%
30	Qwen3.5 27B (Reasoning)	Alibaba	75.6%
31	Gemma 4 31B (Reasoning)	Google	75.6%
32	GPT-5.2 (xhigh)	OpenAI	75.4%
33	GPT-5 mini (high)	OpenAI	75.4%
34	GPT-5.3 Codex (xhigh)	OpenAI	75.4%
35	Qwen: Qwen3.6 Plus	Alibaba	75.2%
36	Gemini 3.5 Flash (medium)	Google	74.6%
37	GPT-5 Codex (high)	OpenAI	74.1%
38	Cohere: Command A	Cohere	73.9%
39	GPT-5.4 (xhigh)	OpenAI	73.9%
40	Gemma 4 12B (Reasoning)	Google	73.5%
41	DeepSeek V4 Flash 0420 (Reasoning, High Effort)	DeepSeek	73.5%
42	GLM-5.2 (max)	Z.ai	73.3%
43	GPT-5.4 mini (xhigh)	OpenAI	73.3%
44	Z.ai: GLM 5 TurboRetiring	Z.ai	73.2%
45	GPT-5 (high)	OpenAI	73.1%
46	GPT-5.1 (high)	OpenAI	72.9%
47	GPT-5.6 Sol (max)	OpenAI	72.7%
48	Qwen3.5 35B A3B (Reasoning)	Alibaba	72.5%
49	Gemma 4 26B A4B (Reasoning)	Google	72.4%
50	GLM-5 (Reasoning)	Z.ai	72.3%

Access this data

JSON GET /api/v1

[
  {
    "rank": 1,
    "model": "Grok 4.3 (medium)",
    "provider": "SpaceXAI",
    "score": "83.3333333333333"
  },
  {
    "rank": 2,
    "model": "Grok 4.20 0309 (Reasoning)",
    "provider": "SpaceXAI",
    "score": "82.9251700680272"
  },
  {
    "rank": 3,
    "model": "MiniMax: MiniMax M3",
    "provider": "MiniMax",
    "score": "82.8571428571429"
  }
]

Curl

$ curl -s https://rightrank.ai/api/v1/rankings/ifbench.json | jq '.[:3]'

CLI brew install rightrank

$ rightrank top ifbench --limit 5
$ rightrank compare claude-opus-4.6 gpt-5.4-pro --category ifbench
$ rightrank value ifbench  # best performance per dollar

Claude Code Skill Skill

# In Claude Code, use the RightRank skill:
/rightrank best model for ifbench
/rightrank compare top 5 by value
/rightrank cheapest model scoring above 80 on ifbench