Skip to content
Get my AI grade

Model efficiency

LiveBench model cost efficiency

45 models ranked by LiveBench cost per successful task.
Recorded Jun 25, 2026Source: LiveBenchHow this board is measuredCSV / JSON

$0.016

DeepSeek V4 Flash ranks first

LiveBench cost per successful task

45

Models ranked

Jun 25

Snapshot recorded

2026

Follow changes

All 45 models by LiveBench cost per successful task

DeepSeek V4 Flash is lowest at $0.016; the middle of the board sits at $0.24.

Highest published row
$1.573
Middle of the board
$0.24
Highest $1.573Middle $0.24
1. DeepSeek V4 Flash: $0.0162. Grok Build 0.1: $0.0243. DeepSeek V4 Pro: $0.0504. MiniMax M3: $0.0605. Grok 4.3: $0.0616. Gemini 3.5 Flash Lite High: $0.0697. GPT 5.4 Nano xHigh: $0.0918. Kimi K2.7 Code: $0.1009. GPT 5.6 Luna xHigh: $0.12310. Grok 4.5: $0.12811. Muse Spark 1.1 High: $0.13812. Kimi K2.6 Thinking: $0.16913. Qwen3.7 Max: $0.18214. GPT 5.2 Codex: $0.18715. GLM 5.2: $0.19516. Qwen3.6 27b: $0.20217. GPT 5.6 Luna Max: $0.20218. GPT 5.6 Terra xHigh: $0.20819. GPT 5.5: $0.21220. Qwen3.6 Plus: $0.22721. GPT 5.2 2025 12 11 High: $0.23422. Muse Spark 1.1 xHigh: $0.23423. Gemini 3.6 Flash High: $0.23524. Gemini 3.5 Flash High: $0.24925. Gemini 3.1 Pro Preview High: $0.26226. Claude Sonnet 4 6 Thinking Auto Medium Effort: $0.30627. GPT 5.4 Mini xHigh: $0.33428. Inkling xHigh: $0.34329. GPT 5.6 Sol xHigh: $0.35530. GPT 5.5 High: $0.36731. Claude Opus 5 High Effort: $0.37332. Kimi K3: $0.37933. GPT 5.4 xHigh: $0.38734. Claude Opus 4 6 Thinking Auto High Effort: $0.40435. Claude Opus 5 xHigh Effort: $0.48736. Claude Sonnet 5 xHigh Effort: $0.49237. GPT 5.6 Terra Max: $0.49738. Claude Opus 4 7 xHigh Effort: $0.52839. GPT 5.5 xHigh: $0.53040. GPT 5.6 Sol Max: $0.58941. Claude Opus 4 5 20251101 Thinking 64K High Effort: $0.61042. Claude Opus 5 Max Effort: $0.64143. Claude Opus 4 8 xHigh Effort: $0.68844. Claude Fable 5 xHigh Effort: $0.81145. Claude Fable 5 Max Effort: $1.573
Rank 1One bar per published rowRank 45

Official LiveBench release: 2026-06-25 · 7 checks behind each row · 100% confidence across rows.

LiveBench model cost efficiency current rankings
RankNameLiveBench cost per successful taskLivebench average input tokensLivebench average output tokensLivebench cost per questionWhy
01DeepSeek V4 Flash$0.016148.8K34.4K$0.01This rank
02Grok Build 0.1$0.02458.8K980$0.02This rank
03DeepSeek V4 Pro$0.05096.6K35K$0.04This rank
04MiniMax M3$0.060331.9K15.5K$0.04This rank
05Grok 4.3$0.06128.7K11.2K$0.04This rank
06Gemini 3.5 Flash Lite High$0.069112.1K11.5K$0.04This rank
07GPT 5.4 Nano xHigh$0.091180.3K46.2K$0.06This rank
08Kimi K2.7 Code$0.100138.1K13K$0.07This rank
09GPT 5.6 Luna xHigh$0.12345.3K13.1K$0.09This rank
10Grok 4.5$0.12861.2K10.5K$0.1This rank
11Muse Spark 1.1 High$0.138296.7K13.8K$0.1This rank
12Kimi K2.6 Thinking$0.16978K27K$0.12This rank
13Qwen3.7 Max$0.18260.7K12.9K$0.13This rank
14GPT 5.2 Codex$0.18737K9,154$0.14This rank
15GLM 5.2$0.195229.4K23.5K$0.14This rank
16Qwen3.6 27b$0.20269.3K24.3K$0.13This rank
17GPT 5.6 Luna Max$0.202126K21.8K$0.15This rank
18GPT 5.6 Terra xHigh$0.20877.4K8,228$0.15This rank
19GPT 5.5$0.21223.7K4,151$0.15Showing
20Qwen3.6 Plus$0.227143.9K28.2K$0.16This rank
21GPT 5.2 2025 12 11 High$0.23450.8K11.5K$0.17This rank
22Muse Spark 1.1 xHigh$0.234200.1K13.7K$0.18This rank
23Gemini 3.6 Flash High$0.235188.1K14.7K$0.17This rank
24Gemini 3.5 Flash High$0.249108.3K15.5K$0.19This rank
25Gemini 3.1 Pro Preview High$0.26299.1K13.4K$0.2This rank
26Claude Sonnet 4 6 Thinking Auto Medium Effort$0.3061,32813.2K$0.22This rank
27GPT 5.4 Mini xHigh$0.33470.7K47.4K$0.22This rank
28Inkling xHigh$0.343355.6K22.2K$0.25This rank
29GPT 5.6 Sol xHigh$0.35586.8K7,161$0.28This rank
30GPT 5.5 High$0.36762.6K7,718$0.29This rank
31Claude Opus 5 High Effort$0.3731,2658,374$0.29This rank
32Kimi K3$0.379261.8K13.6K$0.3This rank
33GPT 5.4 xHigh$0.38777K18.3K$0.3This rank
34Claude Opus 4 6 Thinking Auto High Effort$0.4041,9359,651$0.3This rank
35Claude Opus 5 xHigh Effort$0.4871,24811.1K$0.39This rank
36Claude Sonnet 5 xHigh Effort$0.4921,21419.4K$0.37This rank
37GPT 5.6 Terra Max$0.497179.1K22.1K$0.4This rank
38Claude Opus 4 7 xHigh Effort$0.5281,28511.5K$0.4This rank
39GPT 5.5 xHigh$0.53095.9K11.4K$0.42This rank
40GPT 5.6 Sol Max$0.589181.2K11.7K$0.49This rank
41Claude Opus 4 5 20251101 Thinking 64K High Effort$0.61079K15.2K$0.44This rank
42Claude Opus 5 Max Effort$0.6411,26314.8K$0.51This rank
43Claude Opus 4 8 xHigh Effort$0.6881,25917.1K$0.54This rank
44Claude Fable 5 xHigh Effort$0.8111,34710.5K$0.64This rank
45Claude Fable 5 Max Effort$1.5731,26420.3K$1.27This rank

Why this rank

GPT 5.5 on LiveBench Model Cost Efficiency

Snapshot of Jun 25, 2026, 11:59 PM UTC · Link to this snapshot

Some explanation inputs are unavailable. Not published: the evidence frozen with this snapshot, comparable prior ranking.

Ranks exact model configurations by official cost per successful task for the complete frozen LiveBench workload.

19

Rank on this board

0.212

Score

7

Inputs behind the score

100% confidence · 1 of 1 components evidenced

What the score is built from

1 of 1 evidenced

Every component evidenced · 1 evidence record each

Components of GPT 5.5's score
ComponentContribution
Livebench cost per successful task0.212

New in this snapshot

No change in rank

no_prior_published_snapshot

No component moved since the previous snapshot.

Where each number comes from

7 public records
  • LiveBench average input tokensLivebench average input tokens · observed Jun 25, 2026, 11:59 PM UTC23,651 tokens100% confidence
    Source
    LiveBench official model benchmark
    Grade
    A
    Freshness
    fresh
    Locator
    {"formula":"(sum task cost / sum question count) / (overall score / 100)","costAsset":"https://livebench.ai/cost_2026_06_25.csv","scoreAsset":"https://livebench.ai/table_2026_06_25.csv","categoryAsset":"https://livebench.ai/categories_2026_06_25.json"}
    Open the source
  • LiveBench average output tokensLivebench average output tokens · observed Jun 25, 2026, 11:59 PM UTC4,151 tokens100% confidence
    Source
    LiveBench official model benchmark
    Grade
    A
    Freshness
    fresh
    Locator
    {"formula":"(sum task cost / sum question count) / (overall score / 100)","costAsset":"https://livebench.ai/cost_2026_06_25.csv","scoreAsset":"https://livebench.ai/table_2026_06_25.csv","categoryAsset":"https://livebench.ai/categories_2026_06_25.json"}
    Open the source
  • LiveBench cost per questionLivebench cost per question · observed Jun 25, 2026, 11:59 PM UTC$0.15100% confidence
    Source
    LiveBench official model benchmark
    Grade
    A
    Freshness
    fresh
    Locator
    {"formula":"(sum task cost / sum question count) / (overall score / 100)","costAsset":"https://livebench.ai/cost_2026_06_25.csv","scoreAsset":"https://livebench.ai/table_2026_06_25.csv","categoryAsset":"https://livebench.ai/categories_2026_06_25.json"}
    Open the source
  • LiveBench cost per successful taskLivebench cost per successful task · observed Jun 25, 2026, 11:59 PM UTC$0.21100% confidence
    Source
    LiveBench official model benchmark
    Grade
    A
    Freshness
    fresh
    Locator
    {"formula":"(sum task cost / sum question count) / (overall score / 100)","costAsset":"https://livebench.ai/cost_2026_06_25.csv","scoreAsset":"https://livebench.ai/table_2026_06_25.csv","categoryAsset":"https://livebench.ai/categories_2026_06_25.json"}
    Open the source
  • LiveBench input priceLivebench input price per million · observed Jun 25, 2026, 11:59 PM UTC$5.00100% confidence
    Source
    LiveBench official model benchmark
    Grade
    A
    Freshness
    fresh
    Locator
    {"formula":"(sum task cost / sum question count) / (overall score / 100)","costAsset":"https://livebench.ai/cost_2026_06_25.csv","scoreAsset":"https://livebench.ai/table_2026_06_25.csv","categoryAsset":"https://livebench.ai/categories_2026_06_25.json"}
    Open the source
  • LiveBench output priceLivebench output price per million · observed Jun 25, 2026, 11:59 PM UTC$30.00100% confidence
    Source
    LiveBench official model benchmark
    Grade
    A
    Freshness
    fresh
    Locator
    {"formula":"(sum task cost / sum question count) / (overall score / 100)","costAsset":"https://livebench.ai/cost_2026_06_25.csv","scoreAsset":"https://livebench.ai/table_2026_06_25.csv","categoryAsset":"https://livebench.ai/categories_2026_06_25.json"}
    Open the source
  • LiveBench overall scoreLivebench overall score · observed Jun 25, 2026, 11:59 PM UTC72.154 percent100% confidence
    Source
    LiveBench official model benchmark
    Grade
    A
    Freshness
    fresh
    Locator
    {"formula":"(sum task cost / sum question count) / (overall score / 100)","costAsset":"https://livebench.ai/cost_2026_06_25.csv","scoreAsset":"https://livebench.ai/table_2026_06_25.csv","categoryAsset":"https://livebench.ai/categories_2026_06_25.json"}
    Open the source
LiveBench 2026-06-25 model cost efficiency, version livebench-2026-06-25-official-v1-livebench-model-efficiency-v1315 observations · 1 source · quality passed
Snapshot
9f985b73-8095-4473-9445-5dd484a1ae8f
Data hash
299a87f51afdb539381eb1329d49abaca9dc5bf2659659929ce499a0fb6a671c
Method hash
1cafe5c9d864ba7cdfd0858525b79ab5222403edf79a3fbb8f3dee724546c4cb

More in Model efficiency

Capability relative to token price, context, latency, and throughput.

3 boards

See how often AI names you.

Enter your website. In about two minutes, Rank.ai asks ChatGPT, Claude and Gemini 12 questions your customers ask and grades how often they name you.

  • Your grade out of 100How often AI names you, cites your site, and how high it ranks you.
  • Who gets namedEvery competitor in the answers, most named first.
  • The pages AI readsThe sources behind each answer.
  • Three fixesWhat to fix first, with a brief for the first article.