Best value LLM APIs: quality per dollar
Cheap is easy; cheap and good is the real question. This ranks 64 chat models by a single derived number — their LMArena (Chatbot Arena) Elo divided by their cheapest blended price per 1M tokens — so you see how much measured quality each dollar buys. Each model links to its cheapest provider. All figures computed from the index; Elo sourced here.
Ranked by value (Elo per dollar)
| # | Model | Elo | Cheapest /1M | Verified | Elo / $ | |
|---|---|---|---|---|---|---|
| 1 | glm-5.3-flash · StreamLake | 1475 | $0.189 | today | 7,825 | |
| 2 | deepseek-v4-flash · Relace | 1436 | $0.21 | today | 6,838 | |
| 3 | gemma-4-26b-a4b · Reka | 1438 | $0.26 | 23d ago | 5,531 | |
| 4 | gemma-4-31b · Reka | 1451 | $0.38 | 24d ago | 3,818 | |
| 5 | glm-4.7-flash · Venice | 1366 | $0.46 | 4d ago | 2,970 | |
| 6 | nemotron-3-super-120b-a12b · DeepInfra | 1361 | $0.485 | today | 2,806 | |
| 7 | deepseek-v3.2 · GMICloud | 1425 | $0.518 | today | 2,749 | |
| 8 | minimax-m2.7 · GMICloud | 1415 | $1.05 | today | 1,348 | |
| 9 | minimax-m3 · CoreWeave | 1443 | $1.19 | today | 1,213 | |
| 10 | glm-4.6v · Novita AI | 1378 | $1.2 | today | 1,148 | |
| 11 | minimax-m2.5 · Venice | 1391 | $1.22 | today | 1,140 | |
| 12 | minimax-m2 · Minimax | 1346 | $1.275 | today | 1,056 | |
| 13 | minimax-m2.1 · Novita AI | 1384 | $1.5 | 15d ago | 923 | |
| 14 | mistral-large · Mistral AI | 1414 | $2 | 50d ago ⚠ | 707 | |
| 15 | glm-4.7 · DeepInfra | 1442 | $2.15 | today | 671 | |
| 16 | glm-5.3 · Inceptron | 1483 | $2.36 | today | 628 | |
| 17 | qwen3.5-27b · SiliconFlow | 1408 | $2.25 | today | 626 | |
| 18 | glm-5.2 · DeepInfra | 1472 | $2.362 | today | 623 | |
| 19 | kimi-k2.6 · Chutes | 1461 | $2.4 | today | 609 | |
| 20 | qwen3.5-122b-a10b · SiliconFlow | 1417 | $2.34 | today | 606 | |
| 21 | deepseek-v4 · Relace | 1458 | $2.5 | today | 583 | |
| 22 | glm-5 · GMICloud | 1458 | $2.52 | today | 579 | |
| 23 | kimi-k2.5 · DeepInfra | 1451 | $2.7 | 46d ago ⚠ | 537 | |
| 24 | qwen3.5-397b-a17b · DeepInfra | 1441 | $3.45 | today | 418 | |
| 25 | glm-5.1 · StreamLake | 1466 | $4.002 | today | 366 | |
| 26 | gemini-3.8-flash · Google | 1493 | $4.5 | 6d ago | 332 | |
| 27 | grok-4.6 · xAI | 1456 | $8 | 6d ago | 182 | |
| 28 | mistral-medium-3-5 · Mistral AI | 1427 | $9 | today | 159 | |
| 29 | kimi-k3 · Chutes | 1485 | $11.35 | today | 131 | |
| 30 | gemini-4-argon · Google | 1525 | $12 | 6d ago | 127 | |
| 31 | gpt-5.6-terra · OpenAI | 1466 | $14 | today | 105 | |
| 32 | gpt-5.6-sol · OpenAI | 1483 | $24 | today | 62 | |
| 33 | claude-opus-5 · Anthropic | 1487 | $30 | today | 50 | |
| 34 | claude-fable-5 · Anthropic | 1507 | $60 | today | 25 | |
| 35 | claude-fable-5-1 · Anthropic | 1498 | $60 | 48d ago ⚠ | 25 | |
| 36 | gpt-6-astra · OpenAI | 1480 | $60 | 6d ago | 25 |
Value = Elo ÷ (input + output price per 1M tokens). Since a model's quality is identical across providers, the cheapest host gives the best value — that's the price shown. Models released within the last 12 months; older ones below. Blended is a coarse ranking aid, not a workload estimate; size your own on each model's page.
Older models — released more than 12 months ago (28)
Still priced daily and often still the pragmatic choice, but no longer the current generation.
| # | Model | Elo | Cheapest /1M | Age | Elo / $ |
|---|---|---|---|---|---|
| 37 | gpt-oss-20b · AkashML | 1317 | $0.12 | 14 mo | 10,975 |
| 38 | gpt-oss-120b · Venice | 1352 | $0.18 | 14 mo | 7,511 |
| 39 | gemma-3-27b · DeepInfra | 1365 | $0.24 | 19 mo | 5,688 |
| 40 | qwen3-30b-a3b · StreamLake | 1327 | $0.241 | 17 mo | 5,504 |
| 41 | qwen3-32b · DeepInfra | 1347 | $0.36 | 17 mo | 3,742 |
| 42 | llama-4-scout · DeepInfra | 1321 | $0.4 | 18 mo | 3,302 |
| 43 | llama-3.3-70b · DeepInfra | 1317 | $0.42 | 22 mo | 3,136 |
| 44 | llama-3.2-3b · Parasail | 1166 | $0.38 | 24 mo | 3,068 |
| 45 | gpt-4o-mini · OpenAI | 1317 | $0.75 | 27 mo | 1,756 |
| 46 | qwen-2.5-72b · DeepInfra | 1303 | $0.76 | 25 mo | 1,714 |
| 47 | command-r · Cohere | 1226 | $0.75 | 25 mo | 1,635 |
| 48 | llama-3.1-70b · DeepInfra | 1293 | $0.8 | 27 mo | 1,616 |
| 49 | glm-4.5-air · Novita AI | 1373 | $0.98 | 14 mo | 1,401 |
| 50 | llama-4-maverick · DeepInfra | 1327 | $1 | 18 mo | 1,327 |
| 51 | deepseek-chat-v3 · DeepInfra | 1396 | $1.14 | 18 mo | 1,225 |
| 52 | deepseek-chat-v3.1 · DeepInfra | 1417 | $1.2 | 14 mo | 1,181 |
| 53 | deepseek-v3.1-terminus · SiliconFlow | 1415 | $1.27 | 13 mo | 1,114 |
| 54 | qwen3-coder · DeepInfra | 1387 | $1.3 | 15 mo | 1,067 |
| 55 | qwen3-next-80b-a3b-thinking · Google | 1369 | $1.35 | 13 mo | 1,014 |
| 56 | gpt-4.1-mini · OpenAI | 1382 | $2 | 18 mo | 691 |
| 57 | glm-4.6 · Venice | 1425 | $2.18 | 12 mo | 654 |
| 58 | glm-4.5v · Novita AI | 1353 | $2.4 | 14 mo | 564 |
| 59 | qwen3-235b-a22b-thinking · Alibaba | 1400 | $2.53 | 14 mo | 553 |
| 60 | deepseek-r1 · DeepInfra | 1398 | $2.65 | 16 mo | 528 |
| 61 | minimax-m1 · Minimax | 1364 | $2.6 | 16 mo | 525 |
| 62 | qwen3-vl-235b-a22b-thinking · Alibaba | 1395 | $4.4 | 12 mo | 317 |
| 63 | gpt-4o · OpenAI | 1443 | $12.5 | 29 mo | 115 |
| 64 | command-a · Cohere | 1354 | $12.5 | 19 mo | 108 |
For contrast: the top of the leaderboard
The highest-quality models by raw Elo — usually not the best value, which is the point.
| Elo rank | Model | Elo | Cheapest /1M | Elo / $ |
|---|---|---|---|---|
| 1 | gemini-4-argon | 1525 | $12 | 127 |
| 2 | claude-fable-5 | 1507 | $60 | 25 |
| 3 | claude-fable-5-1 | 1498 | $60 | 25 |
| 4 | gemini-3.8-flash | 1493 | $4.5 | 332 |
| 5 | claude-opus-5 | 1487 | $30 | 50 |
| 6 | kimi-k3 | 1485 | $11.35 | 131 |
| 7 | glm-5.3 | 1483 | $2.36 | 628 |
| 8 | gpt-5.6-sol | 1483 | $24 | 62 |
| 9 | gpt-6-astra | 1480 | $60 | 25 |
| 10 | glm-5.3-flash | 1475 | $0.189 | 7,825 |
LMArena (Chatbot Arena) · verified 2026-10-23 · leaderboard source · full methodology.
FAQ
What is the best value LLM API right now?
As of 2026-10-23, glm-5.3-flash leads on quality per dollar: 7,825 Elo points per dollar — 1475 LMArena (Chatbot Arena) Elo at $0.189 per 1M tokens (blended input + output) via StreamLake.
How is 'value' calculated here?
Value = a model's LMArena Elo divided by its cheapest blended price per 1M tokens (input + output). A model's quality is identical across providers, so the cheapest host gives the best value — that's the price shown.
Is the highest-Elo model also the best value?
No. The top model by raw Elo is gemini-4-argon at 1525 Elo, but at $12 per 1M its value is only 127 Elo per dollar — well below the value leader glm-5.3-flash. Top quality rarely means top value, which is the whole point of this ranking.