Custom ranking

Make it yours

Decide how much each capability matters. Each weight is multiplied by that factor's score and the products are summed; a factor the source has no value for contributes nothing, and every row says how many of your weighted factors it was measured on.

Capability weights

Set how much each capability matters. Weights are relative; a weight of zero removes an axis entirely.

Agentic: 3Coding: 3Knowledge: 3Reasoning: 3Multimodal & Grounded: 3Instruction Following: 3Math: 3Throughput (runtime): 0

Ranking ruleEach weight is multiplied by that factor’s score, the products are summed, and models are ordered by that total, highest first. A factor the source has no value for contributes nothing to the sum.Throughput is a runtime measurement rather than a capability, so it is rescaled to the same 0–100 range the capability scores use before it is weighted — otherwise a point of throughput weight would not buy the same thing as a point of coding weight. It is normalised across the models that pass the SLA thresholds below, so changing a threshold changes the scale.Because a missing factor adds nothing, a model measured on fewer axes totals less than an equally capable model measured on all of them. Each row therefore carries a derived marker saying how many of the weighted factors it had an observed value for; the tooltip names what was missing and the weakest status among the rest.

Runtime SLA thresholds

A model with no runtime measurement cannot satisfy a threshold, so an active threshold excludes it rather than assuming it passes.

Maximum time to first token (seconds)Minimum throughput (tok/s)

Models passing the filters340 of 340

Score versus cost frontier

Catalog score against the 75% input / 25% output blended token price, limited to the models passing your filters. Cost is logarithmic by default; untick for a linear axis.

$ per successful taskLog cost scale

Cheapest first

The ten lowest blended prices among the models your weights rank.

76 of these rows are marked self-hosted by the source: open-weight models with no first-party hosted token rate. 12 of them show a hosted rate matched from another provider; the rest report an unavailable price rather than zero.

Your ranking

230 ranked by weighted total · 110 with no value on any weighted factor

Compare 0/4

Evidence key: Derived

Models ranked by your capability weights
Rank	Model	Weighted total	Factors	Blended $/1M	TTFT	Throughput
1	Claude Fable 5Anthropic · Proprietary	1744.8	7/7	$20.00	100.13 s	68 tok/s
2	GPT-5.6 SolOpenAI · Proprietary	1743.3	7/7	$8.00	100.66 s	94 tok/s
3	Qwen3.8 MaxAlibaba · Open Weight	1731.6	7/7	Unavailable	53.6 s	40 tok/s
4	Claude Opus 4.8Anthropic · Proprietary	1651.5	7/7	$10.00	35.96 s	65 tok/s
5	GPT-5.5OpenAI · Proprietary	1644.9	7/7	$11.25	37.05 s	98 tok/s
6	GPT-5.6 TerraOpenAI · Proprietary	1634.4	7/7	$4.50	Unavailable	Unavailable
7	Claude Opus 5Anthropic · Proprietary	1607.1	6/7	$10.00	59.73 s	62 tok/s
8	GPT-6 AstraOpenAI · Proprietary	1575.3	6/7	$20.00	265.1 s	66 tok/s
9	GPT-5.4OpenAI · Proprietary	1544.1	7/7	$5.63	157.51 s	89 tok/s
10	Claude Opus 5.5Anthropic · Proprietary	1541.4	6/7	$8.00	573.15 s	95 tok/s
11	Claude Sonnet 5.5Anthropic · Proprietary	1540.2	6/7	$4.00	394.99 s	136 tok/s
12	Gemini 3.5 FlashGoogle · Proprietary	1527.3	7/7	$3.38	12.96 s	188 tok/s
13	Qwen3.7 PlusAlibaba · Proprietary	1511.7	7/7	Unavailable	37.25 s	57 tok/s
14	GPT-6 SolOpenAI · Proprietary	1464.3	6/7	$4.00	Unavailable	64 tok/s
15	Qwen3.6 PlusAlibaba · Proprietary	1455.9	7/7	Unavailable	100.68 s	56 tok/s
16	Grok 4.3xAI · Proprietary	1434.3	7/7	$1.56	24.18 s	105 tok/s
17	TMInklingThinking Machines Lab · Open Weight	1430.7	7/7	$2.57	10.7 s	223 tok/s
18	TMInkling-SmallThinking Machines Lab · Open Weight	1422.3	7/7	$0.79	Unavailable	209 tok/s
19	Gemini 3.7 FlashGoogle · Proprietary	1402.8	6/7	$1.50	8.52 s	285 tok/s
20	GPT-5.3 CodexOpenAI · Proprietary	1399.8	6/7	$4.81	54.24 s	122 tok/s
21	Qwen3.7 MaxAlibaba · Proprietary	1395.9	6/7	Unavailable	14.47 s	197 tok/s
22	Muse SparkMeta · Proprietary	1390.8	7/7	Unavailable	Unavailable	Unavailable
23	Gemini 3.8 FlashGoogle · Proprietary	1386.0	6/7	$1.50	26.25 s	149 tok/s
24	GPT-5.2OpenAI · Proprietary	1374.3	7/7	$4.81	81.31 s	80 tok/s
25	Muse Glimmer 30BMeta · Open WeightSelf-hosted	1365.0	7/7	$0.52	Unavailable	103 tok/s

Time to first token: latency to the first answer chunk (Artificial Analysis, via BenchLM). Reasoning models include thinking time, so values can run to tens or hundreds of seconds.

Prices shown are the 75% input / 25% output blend of the source input and output rates — for example Claude Fable 5 at $10.00 in and $50.00 out.

Loading page…