AI coding models compared: intelligence, speed and cost
42 models on the AA Intelligence Index v5.0, with output speed, time to first token, context window, the cheapest per-token price and the measured cost per task. Gemini 4 Carbon scores highest at 59. Last verified 2026-11-01.
| Model & Vendor ↕ | AA Intelligence ↕ | Median Output Speed ↕ | First Answer Token ↕ | Context Window ↕ | Open Weights ↕ | Cheapest Per-Token Price | AA Cost / Task ↕ | Recommended Effort | Sources & Date |
|---|---|---|---|---|---|---|---|---|---|
| Claude Opus 5.5 Anthropic | 58 | 95 t/s | 573 s | 1M | Proprietary |
$4.00 / $20.00 per M
Direct / OpenRouter
|
$10.04 | xhighBudget: medium, 30.1 for $1.70 | Benchmarks Pricing 2026-10-28 |
| Claude Sonnet 5.5 Anthropic | 54 | 134 t/s | 389 s | 1M | Proprietary |
$2.00 / $10.00 per M
Direct / OpenRouter
|
$10.02 | Use Claude Opus 5.5 xhighA cheaper Anthropic model matches every level. Instead of max, Claude Opus 5.5 xhigh scores 43.38 for $5.18 (26% of the cost). | Benchmarks Pricing 2026-10-28 |
| Claude Fable 5.1 Anthropic | 50 | 69 t/s | 191 s | 1M | Proprietary |
$10.00 / $50.00 per M
Direct / OpenRouter
|
$10.89 | Use Claude Opus 5.5 highA cheaper Anthropic model matches every level. Instead of xhigh, Claude Opus 5.5 high scores 34.53 for $2.42 (28% of the cost). | Benchmarks Pricing 2026-10-28 |
| Claude Haiku 5.5 Anthropic | 36 | 232 t/s | 258 s | 1M | Proprietary |
$0.10 / $0.50 per M
Direct / OpenRouter
|
$0.41 | maxBudget: high, 18.64 for $0.14 | Benchmarks Pricing 2026-10-28 |
| GPT-6 Astra OpenAI | 57 | 66 t/s | 265 s | 1M | Proprietary |
$10.00 / $50.00 per M
Direct / OpenRouter
|
$6.24 | xhigh | Benchmarks Pricing 2026-10-28 |
| GPT-6.1 Sol OpenAI | 55 | 55 t/s | 244 s | 1M | Proprietary |
$2.00 / $10.00 per M
Direct / OpenRouter
|
$1.27 | maxBudget: low, 25.6 for $0.34 | Benchmarks Pricing 2026-10-28 |
| GPT-6.1 Terra OpenAI | 51 | 90.9 t/s | 134 s | 1M | Proprietary |
$1.00 / $5.00 per M
Direct (OpenAI API)
|
$0.96 | — | Benchmarks Pricing 2026-10-28 |
| GPT-6 Luna OpenAI | 28 | 124 t/s | 92.88 s | 1M | Proprietary |
$0.10 / $0.50 per M
Direct / OpenRouter
|
$0.15 | maxBudget: medium, 15.97 for $0.09 | Benchmarks Pricing 2026-10-28 |
| GPT-5.6 Terra OpenAI | — not yet scored on AA Index v5.0 | 135 t/s | 132 s | 1M | Proprietary |
$2.00 / $12.00 per M
Direct / OpenRouter
|
— | — | Benchmarks Pricing 2026-10-28 |
| GPT-5.3 Codex OpenAI | — not yet scored on AA Index v5.0 | 124 t/s | 54.24 s | 400k | Proprietary |
$1.75 / $14.00 per M
Direct / OpenRouter
|
— | — | Benchmarks Pricing 2026-10-28 |
| Gemini 4 Argon Google | 50 | 84 t/s | 48.55 s | 1M | Proprietary |
$2.00 / $10.00 per M
Direct / OpenRouter
|
$3.84 | lowVendor swap: GPT-6 Luna max scores 21.87 for $0.17 (4% of the cost) | Benchmarks Pricing 2026-10-28 |
| Gemini 4 Carbon Google | 59 | 99.8 t/s | 97.48 s | 1M | Proprietary |
$2.00 / $10.00 per M
Google (restricted access)
|
$7.76 | medium | Benchmarks Pricing 2026-10-28 |
| Gemini 3.8 Flash Google | 31 | 144 t/s | 26.19 s | 1M | Proprietary |
$0.75 / $3.75 per M
Direct / OpenRouter
|
$7.77 | — | Benchmarks Pricing 2026-10-28 |
| Gemini 3.1 Pro Google | — not yet scored on AA Index v5.0 | 132 t/s | 20.8 s | 1M | Proprietary |
$2.00 / $12.00 per M
Direct / OpenRouter
|
— | — | Benchmarks Pricing 2026-10-28 |
| Grok 4.7 SpaceXAI | 35 | 94 t/s | 44.65 s | 500k | Proprietary |
$2.00 / $6.00 per M
Direct / OpenRouter
|
$6.18 | mediumBudget: low, 20.8 for $3.87Vendor swap: GPT-6.1 Sol low scores 25.6 for $0.34 (4% of the cost) | Benchmarks Pricing 2026-10-28 |
| Kimi K3 Moonshot AI | — not yet scored on AA Index v5.0 | 42 t/s | 3.67 s | 1.05M | Open |
$0.30 / $11.94 per M
OpenRouter
|
— | highBudget: low, 11.1 for $2.98Vendor swap: GPT-6 Luna xhigh scores 18.44 for $0.13 (3% of the cost) | Benchmarks Pricing 2026-10-28 |
| Kimi K2.7 Code Moonshot AI | — not yet scored on AA Index v5.0 | 59 t/s | 2.93 s | 256k | Open |
$0.67 / $3.35 per M
OpenRouter
|
— | — | Benchmarks Pricing 2026-10-28 |
| GLM-5.3 Z.ai | 32 | 86 t/s | 2.44 s | 1M | Open |
$0.079 / $4.80 per M
OpenRouter
|
$5.01 | highVendor swap: GPT-6 Luna medium scores 15.97 for $0.09 (2% of the cost) | Benchmarks Pricing 2026-10-28 |
| GLM-5.3 Flash Z.ai | — not yet scored on AA Index v5.0 | 78 t/s | 2.57 s | 1M | Open |
$0.044 / $0.56 per M
OpenRouter
|
— | maxBudget: high, 12.41 for $0.58Vendor swap: GPT-6 Luna medium scores 15.97 for $0.09 (12% of the cost) | Benchmarks Pricing 2026-10-28 |
| Qwen3.8 Max Alibaba | 34 | 39 t/s | 3.06 s | 984k | Proprietary |
$2.00 / $6.00 per M
OpenRouter
|
$10.02 | — | Benchmarks Pricing 2026-10-28 |
| Qwen3.8 2.4T A95B Alibaba | — not yet scored on AA Index v5.0 | 40 t/s | 3.14 s | 262k | Open |
$2.00 / $6.00 per M
OpenRouter
|
— | — | Benchmarks Pricing 2026-10-28 |
| Qwen3.8 Flash Alibaba | — not yet scored on AA Index v5.0 | 55 t/s | 2.07 s | 256k | Open |
$0.15 / $0.47 per M
OpenRouter
|
— | — | Benchmarks Pricing 2026-10-28 |
| Qwen3.8 27B Alibaba | 20 | 45 t/s | 3.93 s | 256k | Open |
$0.24 / $2.20 per M
OpenRouter
|
$3.53 | — | Benchmarks Pricing 2026-10-28 |
| Qwen3.7 Plus Alibaba | — not yet scored on AA Index v5.0 | 58 t/s | 2.14 s | 1M | Open |
$0.32 / $1.28 per M
OpenRouter
|
— | — | Benchmarks Pricing 2026-10-28 |
| DeepSeek V4.1 Flash DeepSeek | 32 | 216 t/s | 1.08 s | 1M | Open |
$0.15 / $0.60 per M
Direct / OpenRouter
|
$0.60 | highBudget: low, 15.96 for $0.21Vendor swap: GPT-6 Luna xhigh scores 18.44 for $0.13 (48% of the cost) | Benchmarks Pricing 2026-10-28 |
| DeepSeek V4 Pro DeepSeek | 23 | 107 t/s | 1.61 s | 1M | Open |
$0.66 / $1.98 per M
Direct / OpenRouter
|
$1.77 | — | Benchmarks Pricing 2026-10-28 |
| DeepSeek V4 Flash Vision DeepSeek | — not yet scored on AA Index v5.0 | 202 t/s | 1.42 s | 1M | Open |
$0.15 / $0.60 per M
Direct / OpenRouter
|
— | — | Benchmarks Pricing 2026-10-28 |
| MiMo V2.6 Pro Xiaomi | 30 | 49 t/s | 3.22 s | 1M | Open |
$0.43 / $0.87 per M
OpenRouter
|
$0.26 | — | Benchmarks Pricing 2026-10-28 |
| MiMo V2.6 Flash Xiaomi | 27 | 54 t/s | 5.19 s | 1M | Open |
$0.14 / $0.28 per M
OpenRouter
|
$0.13 | — | Benchmarks Pricing 2026-10-28 |
| MiniMax M3 MiniMax | 19 | 103 t/s | 1.18 s | 1M | Open |
$0.30 / $1.20 per M
OpenRouter
|
$1.29 | — | Benchmarks Pricing 2026-10-28 |
| Mistral Large 4 Mistral AI | 25 | 72 t/s | 1.59 s | 524k | Proprietary |
$0.68 / $2.09 per M
OpenRouter
|
$5.54 | maxVendor swap: GPT-6 Luna max scores 21.87 for $0.17 (11% of the cost) | Benchmarks Pricing 2026-10-28 |
| Step 5 Preview StepFun | 29 | 92 t/s | 2.55 s | 1M | Open |
$1.00 / $2.70 per M
OpenRouter
|
$2.35 | — | Benchmarks Pricing 2026-10-28 |
| Ling 3.1 Flash InclusionAI | 29 | 210 t/s | 1.87 s | 1M | Open |
$0.075 / $0.23 per M
OpenRouter
|
$3.19 | — | Benchmarks Pricing 2026-10-28 |
| Muse Wonder Meta | 43 | 161 t/s | 33.81 s | 1.05M | Proprietary |
$4.00 / $20.00 per M
OpenRouter
|
$6.95 | mediumBudget: low, 21.21 for $1.84Vendor swap: Claude Haiku 5.5 xhigh scores 22.28 for $0.23 (9% of the cost) | Benchmarks Pricing 2026-10-28 |
| Muse Spark 1.3 Meta | 36 | 201.8 t/s | 39.01 s | 1M | Proprietary |
$1.25 / $4.25 per M
OpenRouter
|
$2.39 | — | Benchmarks Pricing 2026-10-28 |
| Muse Spark Open Meta | 25 | 141 t/s | 0.83 s | 1.05M | Open |
$1.25 / $4.25 per M
OpenRouter
|
$2.02 | — | Benchmarks Pricing 2026-10-28 |
| Inkling Thinking Machines | 12 | 224 t/s | 1.73 s | 1M | Open |
$1.00 / $4.05 per M
OpenRouter
|
$1.35 | — | Benchmarks Pricing 2026-10-28 |
| Inkling Small Thinking Machines | 13 | 189 t/s | 1.79 s | 1M | Open |
$0.45 / $1.20 per M
OpenRouter
|
$0.20 | maxVendor swap: GPT-6 Luna low scores 4.05 for $0.02 (1% of the cost) | Benchmarks Pricing 2026-10-28 |
| Nemotron 3.5 Super VL NVIDIA | 14 | 435 t/s | 0.83 s | 1M | Open |
$0.10 / $0.40 per M
OpenRouter
|
$0.56 | — | Benchmarks Pricing 2026-10-28 |
| Beam 1.0 Reflection | 18 | 344.9 t/s | 0.56 s | 1.05M | Open |
$1.00 / $3.00 per M
OpenRouter
|
$0.28 | maxBudget: high, 1.74 for $0.61Vendor swap: GPT-6 Luna low scores 4.05 for $0.02 (2% of the cost) | Benchmarks Pricing 2026-10-28 |
| Tencent Hy3 Tencent | — not yet scored on AA Index v5.0 | 84 t/s | 2.81 s | 256k | Open |
$0.13 / $0.53 per M
OpenRouter
|
— | — | Benchmarks Pricing 2026-10-28 |
| Kimi K3.1 Pro Moonshot AI | 46 | 41.4 t/s | 2.95 s | 1M | Proprietary |
$2.00 / $10.00 per M
Moonshot AI
|
$3.68 | — | Benchmarks Pricing 2026-10-28 |
AA Intelligence scores are from the Artificial Analysis Intelligence Index v5.0, at each model's highest measured reasoning effort. Recommended effort is the cheapest level within 90% of the model's best FrontierCode 2.0 score, skipping levels a cheaper model from the same vendor matches within a point; see Optimal effort. Models without a v5.0 score are listed with their speed, price and limits, and left out of the intelligence and value comparisons.
Each dot is a model: higher scores better, further right costs more per task. The teal staircase is the best score money can buy at each price; everything in the shaded zone beneath it has a cheaper model that scores at least as well. Scores run from MiMo V2.6 Flash (27) at $0.13 per task to Gemini 4 Carbon (59) at $7.76. Hover or focus a dot for its cheaper alternative.
Cost per task is Artificial Analysis's measured cost of running its Intelligence Index at API list prices. Subscription plans meter usage differently, so treat this as relative value, not your bill. 11 models have no measured cost per task and are not plotted.
Show as table
| Model | AA Index | Cost / Task | Cheapest Model Scoring At Least As High | Price Gap |
|---|---|---|---|---|
| Gemini 4 Carbon best value | 59 | $7.76 | — | — |
| Claude Opus 5.5 | 58 | $10.04 | Gemini 4 Carbon (59, $7.76) | 1.3x |
| GPT-6 Astra best value | 57 | $6.24 | — | — |
| GPT-6.1 Sol best value | 55 | $1.27 | — | — |
| Claude Sonnet 5.5 | 54 | $10.02 | GPT-6.1 Sol (55, $1.27) | 7.9x |
| GPT-6.1 Terra best value | 51 | $0.96 | — | — |
| Gemini 4 Argon | 50 | $3.84 | GPT-6.1 Terra (51, $0.96) | 4.0x |
| Claude Fable 5.1 | 50 | $10.89 | GPT-6.1 Terra (51, $0.96) | 11.3x |
| Kimi K3.1 Pro | 46 | $3.68 | GPT-6.1 Terra (51, $0.96) | 3.8x |
| Muse Wonder | 43 | $6.95 | GPT-6.1 Terra (51, $0.96) | 7.2x |
| Claude Haiku 5.5 best value | 36 | $0.41 | — | — |
| Muse Spark 1.3 | 36 | $2.39 | Claude Haiku 5.5 (36, $0.41) | 5.8x |
| Grok 4.7 | 35 | $6.18 | Claude Haiku 5.5 (36, $0.41) | 15.1x |
| Qwen3.8 Max | 34 | $10.02 | Claude Haiku 5.5 (36, $0.41) | 24.4x |
| DeepSeek V4.1 Flash | 32 | $0.60 | Claude Haiku 5.5 (36, $0.41) | 1.5x |
| GLM-5.3 | 32 | $5.01 | Claude Haiku 5.5 (36, $0.41) | 12.2x |
| Gemini 3.8 Flash | 31 | $7.77 | Claude Haiku 5.5 (36, $0.41) | 19.0x |
| MiMo V2.6 Pro best value | 30 | $0.26 | — | — |
| Step 5 Preview | 29 | $2.35 | MiMo V2.6 Pro (30, $0.26) | 9.0x |
| Ling 3.1 Flash | 29 | $3.19 | MiMo V2.6 Pro (30, $0.26) | 12.3x |
| GPT-6 Luna best value | 28 | $0.15 | — | — |
| MiMo V2.6 Flash best value | 27 | $0.13 | — | — |
| Muse Spark Open | 25 | $2.02 | MiMo V2.6 Flash (27, $0.13) | 15.5x |
| Mistral Large 4 | 25 | $5.54 | MiMo V2.6 Flash (27, $0.13) | 42.6x |
| DeepSeek V4 Pro | 23 | $1.77 | MiMo V2.6 Flash (27, $0.13) | 13.6x |
| Qwen3.8 27B | 20 | $3.53 | MiMo V2.6 Flash (27, $0.13) | 27.2x |
| MiniMax M3 | 19 | $1.29 | MiMo V2.6 Flash (27, $0.13) | 9.9x |
| Beam 1.0 | 18 | $0.28 | MiMo V2.6 Flash (27, $0.13) | 2.2x |
| Nemotron 3.5 Super VL | 14 | $0.56 | MiMo V2.6 Flash (27, $0.13) | 4.3x |
| Inkling Small | 13 | $0.20 | MiMo V2.6 Flash (27, $0.13) | 1.5x |
| Inkling | 12 | $1.35 | MiMo V2.6 Flash (27, $0.13) | 10.4x |
Each line is one model, each dot one effort level: higher scores better on FrontierCode 2.0, further right costs more per task. The teal dot is the recommended level: the cheapest one within 90% of that model's best score, skipping any level a cheaper model from the same vendor matches. A model with no teal dot is outclassed by a cheaper sibling at every level. Hover or focus a dot for what it gives up, and for a cheaper model from another vendor that scores within a point of it.
Scores and costs are from Cognition's FrontierCode 2.0 (checked 2026-10-28): 132 maintainer-written tasks graded on whether the pull request would be merged. Each model ran in an agent harness, its vendor's own where there is one (Claude Code, Codex, Grok Build) and Cognition's Devin CLI otherwise; the table names it. Cost is the mean API spend per run at list prices; subscription plans spend usage limits instead of dollars. The benchmark publishes no error bars, so treat gaps of a point or two as ties. It reports no timings either; tokens per task, in the table, are the closest stand-in for speed. Faded levels in the table are matched for less by a cheaper model from the same vendor.
Show as table
| Model | Score · Cost · Tokens per Effort Level | Why | Reasoning | Harness |
|---|---|---|---|---|
| Claude Opus 5.5 | low 14.72 · $0.75 · 14k tokens medium 30.1 · $1.70 · 30k tokens high 34.53 · $2.42 · 41k tokens xhigh 43.38 · $5.18 · 88k tokens max 45.45 · $12.82 · 202k tokens |
Optimal effort: xhigh Budget effort: medium Vendor swap: — |
low 14.72 · $0.75 → Claude Haiku 5.5 high 18.64 · $0.14 Optimal xhigh 43.38 · $5.18 (95% score · 40% cost of max) Budget medium 30.1 · $1.70 (66% score · 13% cost of max) |
Claude Code |
| GPT-6 Astra | low 31.11 · $2.44 · 9k tokens medium 36.75 · $3.49 · 14k tokens high 38.31 · $4.28 · 18k tokens xhigh 41.41 · $4.54 · 21k tokens max 42.77 · $6.68 · 36k tokens |
Optimal effort: xhigh Budget effort: — Vendor swap: — |
low 31.11 · $2.44 → GPT-6.1 Sol high 34.82 · $0.65 medium 36.75 · $3.49 → GPT-6.1 Sol max 38.95 · $1.03 high 38.31 · $4.28 → GPT-6.1 Sol max 38.95 · $1.03 Optimal xhigh 41.41 · $4.54 (97% score · 68% cost of max) |
Codex |
| Claude Sonnet 5.5 | low 9.41 · $0.37 · 11k tokens medium 14.17 · $0.48 · 14k tokens high 21.9 · $1.07 · 28k tokens xhigh 34.75 · $3.31 · 80k tokens max 40.96 · $19.94 · 515k tokens |
Optimal effort: — Use instead: Claude Opus 5.5 xhigh |
max 40.96 · $19.94 → Claude Opus 5.5 xhigh 43.38 · $5.18 (106% score · 26% cost) | Claude Code |
| Gemini 4 Carbon | medium 40 · $4.04 · 109k tokens high 37.48 · $4.43 · 121k tokens low 31.88 · $4.55 · 58k tokens |
Optimal effort: medium Budget effort: — Vendor swap: — |
Optimal medium 40 · $4.04 | Devin CLI |
| GPT-6.1 Sol | low 25.6 · $0.34 · 10k tokens medium 29.58 · $0.49 · 16k tokens high 34.82 · $0.65 · 22k tokens xhigh 32.33 · $0.69 · 25k tokens max 38.95 · $1.03 · 41k tokens |
Optimal effort: max Budget effort: low Vendor swap: — |
Optimal max 38.95 · $1.03 Budget low 25.6 · $0.34 (66% score · 33% cost of max) |
Codex |
| Claude Fable 5.1 | low 20.59 · $3.01 · 29k tokens medium 21.47 · $3.89 · 38k tokens high 25.67 · $5.39 · 54k tokens xhigh 32.44 · $8.68 · 93k tokens max 31.67 · $11.46 · 119k tokens |
Optimal effort: — Use instead: Claude Opus 5.5 high |
xhigh 32.44 · $8.68 → Claude Opus 5.5 high 34.53 · $2.42 (106% score · 28% cost) | Claude Code |
| Claude Haiku 5.5 | low 10.46 · $0.07 · 39k tokens medium 13.42 · $0.10 · 58k tokens high 18.64 · $0.14 · 81k tokens xhigh 22.28 · $0.23 · 128k tokens max 27.72 · $0.37 · 205k tokens |
Optimal effort: max Budget effort: high Vendor swap: — |
Optimal max 27.72 · $0.37 Budget high 18.64 · $0.14 (67% score · 38% cost of max) |
Claude Code |
| Grok 4.7 | low 20.8 · $3.87 · 38k tokens medium 25.94 · $8.03 · 72k tokens high 26.55 · $11.09 · 94k tokens xhigh 28.62 · $14.36 · 111k tokens |
Optimal effort: medium Budget effort: low Vendor swap: GPT-6.1 Sol low |
Optimal medium 25.94 · $8.03 (91% score · 56% cost of xhigh) Budget low 20.8 · $3.87 (73% score · 27% cost of xhigh) Swap GPT-6.1 Sol low 25.6 · $0.34 (99% score · 4% cost of medium) |
Grok Build |
| Muse Wonder | low 21.21 · $1.84 · 20k tokens medium 23.14 · $2.53 · 27k tokens high 24.28 · $3.14 · 33k tokens xhigh 24.11 · $3.79 · 39k tokens |
Optimal effort: medium Budget effort: low Vendor swap: Claude Haiku 5.5 xhigh |
Optimal medium 23.14 · $2.53 (95% score · 81% cost of high) Budget low 21.21 · $1.84 (87% score · 59% cost of high) Swap Claude Haiku 5.5 xhigh 22.28 · $0.23 (96% score · 9% cost of medium) |
muse |
| GPT-6 Luna | low 4.05 · $0.02 · 7k tokens medium 15.97 · $0.09 · 31k tokens high 16.5 · $0.12 · 42k tokens xhigh 18.44 · $0.13 · 51k tokens max 21.87 · $0.17 · 80k tokens |
Optimal effort: max Budget effort: medium Vendor swap: — |
Optimal max 21.87 · $0.17 Budget medium 15.97 · $0.09 (73% score · 53% cost of max) |
Codex |
| Gemini 4 Argon | low 21.47 · $3.98 · 63k tokens medium 23.45 · $5.69 · 124k tokens high 26.3 · $6.51 · 149k tokens |
Optimal effort: low Budget effort: — Vendor swap: GPT-6 Luna max |
medium 23.45 · $5.69 → Gemini 4 Carbon medium 40 · $4.04 high 26.3 · $6.51 → Gemini 4 Carbon medium 40 · $4.04 Optimal low 21.47 · $3.98 Swap GPT-6 Luna max 21.87 · $0.17 (102% score · 4% cost of low) |
Devin CLI |
| Mistral Large 4 | low 9.5 · $0.67 · 28k tokens high 8.78 · $1.03 · 51k tokens max 19.65 · $1.55 · 88k tokens |
Optimal effort: max Budget effort: — Vendor swap: GPT-6 Luna max |
Optimal max 19.65 · $1.55 Swap GPT-6 Luna max 21.87 · $0.17 (111% score · 11% cost of max) |
Mistral Vibe |
| DeepSeek V4.1 Flash | low 15.96 · $0.21 · 41k tokens high 18.21 · $0.27 · 57k tokens max 19.3 · $0.38 · 82k tokens |
Optimal effort: high Budget effort: low Vendor swap: GPT-6 Luna xhigh |
Optimal high 18.21 · $0.27 (94% score · 71% cost of max) Budget low 15.96 · $0.21 (83% score · 54% cost of max) Swap GPT-6 Luna xhigh 18.44 · $0.13 (101% score · 48% cost of high) |
Devin CLI |
| Kimi K3 | low 11.1 · $2.98 · 29k tokens high 17.74 · $4.47 · 50k tokens max 17.31 · $5.46 · 64k tokens |
Optimal effort: high Budget effort: low Vendor swap: GPT-6 Luna xhigh |
Optimal high 17.74 · $4.47 Budget low 11.1 · $2.98 (63% score · 67% cost of high) Swap GPT-6 Luna xhigh 18.44 · $0.13 (104% score · 3% cost of high) |
Devin CLI |
| GLM-5.3 | low 14.37 · $3.19 · 47k tokens high 16.84 · $5.18 · 73k tokens max 17.81 · $7.52 · 101k tokens |
Optimal effort: high Budget effort: — Vendor swap: GPT-6 Luna medium |
low 14.37 · $3.19 → GLM-5.3 Flash max 14.18 · $0.79 Optimal high 16.84 · $5.18 (95% score · 69% cost of max) Swap GPT-6 Luna medium 15.97 · $0.09 (95% score · 2% cost of high) |
Devin CLI |
| GLM-5.3 Flash | low 7.81 · $0.43 · 35k tokens high 12.41 · $0.58 · 57k tokens max 14.18 · $0.79 · 90k tokens |
Optimal effort: max Budget effort: high Vendor swap: GPT-6 Luna medium |
Optimal max 14.18 · $0.79 Budget high 12.41 · $0.58 (88% score · 74% cost of max) Swap GPT-6 Luna medium 15.97 · $0.09 (113% score · 12% cost of max) |
Devin CLI |
| Inkling Small | high 1.27 · $1.37 · 14k tokens max 3.21 · $3.68 · 39k tokens |
Optimal effort: max Budget effort: — Vendor swap: GPT-6 Luna low |
Optimal max 3.21 · $3.68 Swap GPT-6 Luna low 4.05 · $0.02 (126% score · 1% cost of max) |
Devin CLI |
| Beam 1.0 | minimal 0 · $0.06 · 2k tokens low 0 · $0.14 · 3k tokens medium 0.83 · $0.32 high 1.74 · $0.61 xhigh 2.43 · $1.13 max 2.9 · $1.30 |
Optimal effort: max Budget effort: high Vendor swap: GPT-6 Luna low |
Optimal max 2.9 · $1.30 Budget high 1.74 · $0.61 (60% score · 47% cost of max) Swap GPT-6 Luna low 4.05 · $0.02 (140% score · 2% cost of max) |
Devin CLI |