Codesota · Agentic · OpenRouter modelsInverted OpenRouter viewSnapshot · 2026-04-27
§ Agentic · Model adoption

OpenRouter models,
ranked by real agent usage.

Direct answer: OpenRouter is a model router and catalog. This page ranks OpenRouter models by app adoption, token volume, estimated spend, and top-1 usage so you can choose by coding, price, context length, latency, or breadth instead of reading a flat provider list.

Fig 0 · Concentration

One number explains the whole market.

Of every dollar agents spend on inference, this is who collects it. The top three vendors take more than the rest combined — that's the entire story before any leaderboard.

anthropic
66.3%
$50.42M · 26 apps
xiaomi
10.4%
$7.93M · 17 apps
openai
6.2%
$4.72M · 23 apps
z-ai
4.9%
$3.70M · 21 apps
qwen
3.4%
$2.57M · 22 apps
minimax
3.1%
$2.35M · 18 apps
20 others
5.7%
$4.33M · 132 apps
Of $76.01M total monthly spend, the top three vendors capture 83.0%. Concentration this high is rare — most catalog markets are flatter.
§ 01 · Four winners

One market, four different leaders.

The same 30-day window produces four different #1s depending on what you measure. Dollars, tokens, breadth of adoption, and depth of adoption disagree — that disagreement is the story.

  • Most dollars
    Anthropic: Claude Opus 4.6
    $22.58M/month across 22 apps. Premium pricing compounds even at modest volume.
  • Most tokens
    Xiaomi: MiMo-V2-Pro
    4.92T through 15 apps. High-volume cheap models absorb the long tail.
  • Most adopted
    Qwen: Qwen3.6 Plus
    Shows up in 28 of 30 apps. “Everyone tries it at least a little” is not the same as “anyone uses it primarily”.
  • Most #1 slots
    Xiaomi: MiMo-V2-Pro
    The top model in 4 of 30 apps. The closest proxy for what agents actually lean on when it matters.
§ 02 · Vendors

Rolled up by vendor.

Same data, one level coarser. Every model per vendor, summed into one row. Blended $/M is vendor-weighted by real usage, so premium vendors show their true price per token.

#VendorModelsAppsTokensMonthly costAvg $/M#1 slots
1anthropic11266.34T$50.42M$7.957
2xiaomi3175.39T$7.93M$1.474
3openai1923900.1B$4.72M$5.252
4z-ai6212.03T$3.70M$1.820
5qwen13223.32T$2.57M$0.774
6minimax2184.92T$2.35M$0.483
7google13282.10T$2.35M$1.124
8moonshotai219576.0B$661K$1.151
9deepseek8201.50T$448K$0.304
10stepfun1152.73T$426K$0.160
11nvidia1121.43T$274K$0.190
12arcee-ai27412.3B$105K$0.260
13x-ai4775.6B$42K$0.550
14amazon1119.6B$18K$0.920
15mistralai21184.8B$5K$0.031
16aion-labs11576.4M$590$1.020
17sao10k219.4B$578$0.060
18microsoft11884.4M$548$0.620
19kwaipilot11517.1M$285$0.550
20anthracite-org1141.9M$149$3.560
21thedrummer21204.2M$75$0.370
22nousresearch21150.7M$30$0.200
23gryphe1134.7M$2$0.060
24openrouter18431.2B$00
25inclusionai22312.9B$00
26tencent1465.0B$00
Fig 1 · Vendor rollup. Copper rows mark the top three by monthly cost.
Fig 2 · Cost vs volume

Where every vendor actually sits.

Monthly spend vs token volume, one bubble per vendor. Bubble size scales with apps using that vendor. The diagonal is even pricing — anything above charges more per token than the cluster, anything below charges less.

$1$10$100$1K$10K$100K$1.00M$10.00M$100.00M10.0M100.0M1.0B10.0B100.0B1.00T10.00TTokens / month →↑ Monthly spendanthropicxiaomiopenaiz-aiqwenminimaxgooglemoonshotaideepseekstepfunnvidiaarcee-aix-aiamazonmistralaiaion-labssao10kmicrosoftkwaipilotanthracite-orgthedrummernousresearchgryphe
Both axes are log-scaled. Bubble area scales with app adoption. Vendors in the upper-left charge more per token; lower-right vendors are the volume-cheap workhorses.
§ 03 · By cost

Top 30 by monthly cost.

Where the dollars actually go. Summed across every app that uses each model.

#ModelVendor$/M in$/M outTokensMonthlyApps#1Top-3
1Anthropic: Claude Opus 4.6anthropic$5.00$25.002.13T$22.58M2239
2Anthropic: Claude Sonnet 4.6anthropic$3.00$15.003.09T$19.65M23310
3Xiaomi: MiMo-V2-Proxiaomi$1.00$3.004.92T$7.68M1545
4Anthropic: Claude Opus 4.7anthropic$5.00$25.00510.0B$5.41M1302
5OpenAI: GPT-5.4openai$2.50$15.00520.4B$3.12M1513
6Z.ai: GLM 5 Turboz-ai$1.20$4.001.56T$3.09M301
7Qwen: Qwen3.6 Plusqwen$0.33$1.953.25T$2.53M2838
8Anthropic: Claude Sonnet 4.5anthropic$3.00$15.00267.7B$1.70M1513
9OpenAI: GPT-5.3-Codexopenai$1.75$14.00276.8B$1.43M912
10MiniMax: MiniMax M2.5minimax$0.15$1.153.02T$1.30M1233
11Google: Gemini 3 Flash Previewgoogle$0.50$3.00891.1B$1.07M2034
12MiniMax: MiniMax M2.7minimax$0.30$1.201.91T$1.05M1703
13Google: Gemini 3.1 Pro Preview Custom Toolsgoogle$2.00$12.00141.5B$679K1902
14Anthropic: Claude Haiku 4.5anthropic$1.00$5.00306.0B$649K901
15Z.ai: GLM 5.1z-ai$1.05$3.50255.9B$444K1702
16StepFun: Step 3.5 Flashstepfun$0.10$0.302.73T$426K1503
17Anthropic: Claude Opus 4.5anthropic$5.00$25.0037.5B$397K601
18MoonshotAI: Kimi K2.5moonshotai$0.44$2.00413.8B$363K1400
19DeepSeek: DeepSeek V3.2deepseek$0.25$0.381.23T$355K2045
20MoonshotAI: Kimi K2.6moonshotai$0.74$4.66162.2B$298K1111
21Google: Gemini 2.5 Flashgoogle$0.30$2.50318.9B$292K912
22NVIDIA: Nemotron 3 Supernvidia$0.09$0.451.43T$274K1200
23Xiaomi: MiMo-V2-Omnixiaomi$0.40$2.00257.2B$218K201
24Google: Gemini 2.5 Progoogle$1.25$10.0043.4B$161K700
25Google: Gemini 2.5 Flash Litegoogle$0.10$0.40643.7B$118K602
26Z.ai: GLM 5z-ai$0.60$2.08115.1B$117K1101
27Arcee AI: Trinity Large Previewarcee-ai$0.15$0.45356.9B$84K600
28OpenAI: GPT-5.4 Miniopenai$0.75$4.5040.6B$73K500
29OpenAI: GPT-5.5openai$5.00$30.002.9B$35K300
30Xiaomi: MiMo-V2-Flashxiaomi$0.09$0.29214.2B$31K400
§ 04 · By volume

Top 30 by raw token volume.

How much work each model carries. Cheap models lead here.

#ModelVendor$/M in$/M outTokensMonthlyApps#1Top-3
1Xiaomi: MiMo-V2-Proxiaomi$1.00$3.004.92T$7.68M1545
2Qwen: Qwen3.6 Plusqwen$0.33$1.953.25T$2.53M2838
3Anthropic: Claude Sonnet 4.6anthropic$3.00$15.003.09T$19.65M23310
4MiniMax: MiniMax M2.5minimax$0.15$1.153.02T$1.30M1233
5StepFun: Step 3.5 Flashstepfun$0.10$0.302.73T$426K1503
6Anthropic: Claude Opus 4.6anthropic$5.00$25.002.13T$22.58M2239
7MiniMax: MiniMax M2.7minimax$0.30$1.201.91T$1.05M1703
8Z.ai: GLM 5 Turboz-ai$1.20$4.001.56T$3.09M301
9NVIDIA: Nemotron 3 Supernvidia$0.09$0.451.43T$274K1200
10DeepSeek: DeepSeek V3.2deepseek$0.25$0.381.23T$355K2045
11Google: Gemini 3 Flash Previewgoogle$0.50$3.00891.1B$1.07M2034
12Google: Gemini 2.5 Flash Litegoogle$0.10$0.40643.7B$118K602
13OpenAI: GPT-5.4openai$2.50$15.00520.4B$3.12M1513
14Anthropic: Claude Opus 4.7anthropic$5.00$25.00510.0B$5.41M1302
15openrouter/elephant-alphaopenrouter431.2B$0800
16MoonshotAI: Kimi K2.5moonshotai$0.44$2.00413.8B$363K1400
17Arcee AI: Trinity Large Previewarcee-ai$0.15$0.45356.9B$84K600
18Google: Gemini 2.5 Flashgoogle$0.30$2.50318.9B$292K912
19Anthropic: Claude Haiku 4.5anthropic$1.00$5.00306.0B$649K901
20OpenAI: GPT-5.3-Codexopenai$1.75$14.00276.8B$1.43M912
21Anthropic: Claude Sonnet 4.5anthropic$3.00$15.00267.7B$1.70M1513
22Xiaomi: MiMo-V2-Omnixiaomi$0.40$2.00257.2B$218K201
23Z.ai: GLM 5.1z-ai$1.05$3.50255.9B$444K1702
24inclusionAI: Ling-2.6-1T (free)Freeinclusionai$0.00$0.00231.3B$0200
25Xiaomi: MiMo-V2-Flashxiaomi$0.09$0.29214.2B$31K400
26Mistral: Mistral Nemomistralai$0.02$0.04184.6B$5K111
27MoonshotAI: Kimi K2.6moonshotai$0.74$4.66162.2B$298K1111
28Google: Gemini 3.1 Pro Preview Custom Toolsgoogle$2.00$12.00141.5B$679K1902
29Z.ai: GLM 5z-ai$0.60$2.08115.1B$117K1101
30DeepSeek: DeepSeek V3 0324deepseek$0.20$0.7786.8B$31K302
§ 05 · By breadth

Top 30 by number of apps using it.

Breadth of adoption — models that appear in the top-20 mix of the most apps.

#ModelVendor$/M in$/M outTokensMonthlyApps#1Top-3
1Qwen: Qwen3.6 Plusqwen$0.33$1.953.25T$2.53M2838
2Anthropic: Claude Sonnet 4.6anthropic$3.00$15.003.09T$19.65M23310
3Anthropic: Claude Opus 4.6anthropic$5.00$25.002.13T$22.58M2239
4DeepSeek: DeepSeek V3.2deepseek$0.25$0.381.23T$355K2045
5Google: Gemini 3 Flash Previewgoogle$0.50$3.00891.1B$1.07M2034
6Google: Gemini 3.1 Pro Preview Custom Toolsgoogle$2.00$12.00141.5B$679K1902
7MiniMax: MiniMax M2.7minimax$0.30$1.201.91T$1.05M1703
8Z.ai: GLM 5.1z-ai$1.05$3.50255.9B$444K1702
9Xiaomi: MiMo-V2-Proxiaomi$1.00$3.004.92T$7.68M1545
10StepFun: Step 3.5 Flashstepfun$0.10$0.302.73T$426K1503
11OpenAI: GPT-5.4openai$2.50$15.00520.4B$3.12M1513
12Anthropic: Claude Sonnet 4.5anthropic$3.00$15.00267.7B$1.70M1513
13MoonshotAI: Kimi K2.5moonshotai$0.44$2.00413.8B$363K1400
14Anthropic: Claude Opus 4.7anthropic$5.00$25.00510.0B$5.41M1302
15MiniMax: MiniMax M2.5minimax$0.15$1.153.02T$1.30M1233
16NVIDIA: Nemotron 3 Supernvidia$0.09$0.451.43T$274K1200
17MoonshotAI: Kimi K2.6moonshotai$0.74$4.66162.2B$298K1111
18Z.ai: GLM 5z-ai$0.60$2.08115.1B$117K1101
19Google: Gemini 2.5 Flashgoogle$0.30$2.50318.9B$292K912
20Anthropic: Claude Haiku 4.5anthropic$1.00$5.00306.0B$649K901
21OpenAI: GPT-5.3-Codexopenai$1.75$14.00276.8B$1.43M912
22openrouter/elephant-alphaopenrouter431.2B$0800
23Google: Gemini 2.5 Progoogle$1.25$10.0043.4B$161K700
24Google: Gemini 3.1 Flash Lite Previewgoogle$0.25$1.5039.1B$23K701
25Google: Gemini 2.5 Flash Litegoogle$0.10$0.40643.7B$118K602
26Arcee AI: Trinity Large Previewarcee-ai$0.15$0.45356.9B$84K600
27xAI: Grok 4.1 Fastx-ai$0.20$0.5057.7B$16K600
28Anthropic: Claude Opus 4.5anthropic$5.00$25.0037.5B$397K601
29OpenAI: GPT-5.4 Miniopenai$0.75$4.5040.6B$73K500
30Xiaomi: MiMo-V2-Flashxiaomi$0.09$0.29214.2B$31K400
§ 06 · By depth

Models that rank #1 in an app's mix (14 models).

Strongest 'primary driver' signal. Being #1 in even one app beats showing up in a long tail.

#ModelVendor$/M in$/M outTokensMonthlyApps#1Top-3
1Xiaomi: MiMo-V2-Proxiaomi$1.00$3.004.92T$7.68M1545
2DeepSeek: DeepSeek V3.2deepseek$0.25$0.381.23T$355K2045
3Qwen: Qwen3.6 Plusqwen$0.33$1.953.25T$2.53M2838
4Anthropic: Claude Sonnet 4.6anthropic$3.00$15.003.09T$19.65M23310
5MiniMax: MiniMax M2.5minimax$0.15$1.153.02T$1.30M1233
6Anthropic: Claude Opus 4.6anthropic$5.00$25.002.13T$22.58M2239
7Google: Gemini 3 Flash Previewgoogle$0.50$3.00891.1B$1.07M2034
8OpenAI: GPT-5.4openai$2.50$15.00520.4B$3.12M1513
9Google: Gemini 2.5 Flashgoogle$0.30$2.50318.9B$292K912
10OpenAI: GPT-5.3-Codexopenai$1.75$14.00276.8B$1.43M912
11Anthropic: Claude Sonnet 4.5anthropic$3.00$15.00267.7B$1.70M1513
12Mistral: Mistral Nemomistralai$0.02$0.04184.6B$5K111
13MoonshotAI: Kimi K2.6moonshotai$0.74$4.66162.2B$298K1111
14Qwen: Qwen3 235B A22Bqwen$0.46$1.8215.1B$13K111
§ 07 · Methodology

How the inverted view is built.

Data source, matching rules, and known limitations — in the open.

  • App leaderboard, inverted.
    Built by inverting the app leaderboard JSON — for every (app, model) pair in the top-20 slice, aggregate by model.
  • Permaslug matching.
    Permaslugs with date suffixes (e.g. anthropic/claude-4.6-opus-20260205) are matched back to the base catalog ID via vendor-scoped token-bag overlap, so dollar totals are not missed.
  • What '#1 in N apps' means.
    The model holds the top token slot for that app's 30-day window. It does not mean the model is objectively best — just that this particular app routes the most tokens to it.
  • Free and stealth lanes.
    Free models contribute to token and adoption rankings but $0 to spend. Stealth / alpha lanes without listed pricing are counted in adoption but not dollars.
  • Regenerate locally.
    npx tsx scripts/fetch-openrouter-apps.ts — fetches fresh data and rebuilds the index.
§ Final · Related

What to read next.

We reply within 48 hours

Model getting ignored?

Think a model deserves a closer look or isn't priced correctly? Tell us and we'll re-run the match against our catalog.

Tell us what you found →
No newsletter Real humans read this 30 seconds to send