Every model One bill
conifer

TowardsPrivate Superintelligence

An ASCII luminance dither of a conifer, framed by a sun and moon. Hover with a pointer to ripple the characters along the row.

Routing from the bottom up.

Local-first

Every query starts on your own machine and only moves up when it has to:

  • Your machine is the first stop. The weights stay loaded in unified memory, so most queries run at full speed and cost nothing.
  • A query goes to the cloud only when the local answer falls short. Cheaper cloud models are tried before the frontier ones.
  • You pay only for the queries that reach the cloud. Most never leave the machine, which cuts total token cost by more than 80%.
  • Turn on secure routing and nothing leaves the machine. The router still picks the best local model for the task.

Local inference
shouldn't feel slow.

GPU accelerated

The speed lives in the kernels. We tune the GPU code to each model and each chip:

  • Hand-written Metal and CUDA kernels on the hot paths, not a generic BLAS call.
  • Fused kernels: quantized matmul, RMSNorm, RoPE and the attention softmax run in one pass, so the weights cross the memory bus once.
  • Decode is memory-bound, so we keep weights in Q4_K and move the fewest bytes per token.
  • A flash-style attention kernel sized to each model's head dimension keeps the KV cache on-chip.
  • For mixture-of-experts models, routing is one grouped GEMM instead of a loop over experts.

The chip spends its time computing, not waiting on memory.

Benchmarks

Each model gets a scorecard.

Live intelligence index

The chart is live. Every open model Conifer runs locally is scored on the same tests as the big cloud models. The numbers come from real hardware, not estimates, and the router reads the same scorecard when it decides whether local is enough.

Intelligence

Artificial Analysis Intelligence Index v4.0 · higher is better · Artificial Analysis · 56 models

  1. 65
    Claude Fable 5: 65 index
  2. 61
    Claude Opus 4.8: 61 index
  3. 60
    GPT-5.5: 60 index
  4. 57
    Claude Opus 4.7: 57 index
  5. 54
    Kimi K2.6: 54 index
  6. 52
    DeepSeek V4 Pro: 52 index
  7. 48
    GPT-5.1 high: 48 index
  8. 48
    Gemini 3 Pro: 48 index
  9. 45
    GPT-5: 45 index
  10. 43
    Claude Opus 4.5: 43 index
  11. 42
    Grok 4: 42 index
  12. 41
    Kimi K2 Thinking: 41 index
  13. 37
    Claude Sonnet 4.5: 37 index
  14. 35
    Gemini 2.5 Pro: 35 index
  15. 33.3
    gpt-oss 120B high: 33.3 index
  16. 33
    GLM-4.6: 33 index
  17. 32.1
    DeepSeek V3.2: 32.1 index
  18. 31.4
    Qwen3 Max: 31.4 index
  19. 27
    DeepSeek R1 0528: 27 index
  20. 25
    Qwen3 235B 2507: 25 index
  21. 24.7
    Qwen3 VL 32B: 24.7 index
  22. 24.6
    Nova 2.0 Lite: 24.6 index
  23. 24.5
    gpt-oss 20B: 24.5 index
  24. 24.4
    MiniMax M1: 24.4 index
  25. 24.4
    GPT-5.4 nano: 24.4 index
  26. 24.3
    Gemini 2.5 Flash: 24.3 index
  27. 24.3
    Nemotron 3 Nano: 24.3 index
  28. 23.7
    o1-preview: 23.7 index
  29. 23.6
    Grok 4.1 Fast: 23.6 index
  30. 23.3
    GPT-5.4 mini: 23.3 index
  31. 23.2
    Nova 2.0 Omni: 23.2 index
  32. 23.1
    Grok 4 Fast: 23.1 index
  33. 23.1
    Nova 2.0 Pro: 23.1 index
  34. 23
    GLM-4.5 Air: 23 index
  35. 22.9
    GPT-4.1 mini: 22.9 index
  36. 22.8
    Mistral Large 3: 22.8 index
  37. 22.6
    Qwen3.5 4B: 22.6 index, runs locally on Conifer
  38. 22.4
    Qwen3 30B A3B 2507: 22.4 index, runs locally on Conifer
  39. 22.3
    DeepSeek V3 0324: 22.3 index
  40. 22
    Devstral 2: 22 index
  41. 18.8
    Mistral Medium 3: 18.8 index
  42. 18.4
    Llama 4 Maverick: 18.4 index
  43. 17
    Qwen3 32B: 17 index, runs locally on Conifer
  44. 17
    R1 Distill 32B: 17 index, runs locally on Conifer
  45. 16
    Qwen3 14B: 16 index, runs locally on Conifer
  46. 16
    R1 Distill 14B: 16 index, runs locally on Conifer
  47. 14
    Llama 3.3 70B: 14 index
  48. 14
    Qwen3 4B: 14 index, runs locally on Conifer
  49. 13
    Qwen3 8B: 13 index, runs locally on Conifer
  50. 13
    Qwen2.5 32B: 13 index, runs locally on Conifer
  51. 13
    Qwen2.5 Coder 32B: 13 index, runs locally on Conifer
  52. 12
    Llama 3.1 8B: 12 index, runs locally on Conifer
  53. 10.4
    Phi-4: 10.4 index, runs locally on Conifer
  54. 10.3
    Gemma 3 27B: 10.3 index
  55. 9
    Gemma 3 12B: 9 index, runs locally on Conifer
  56. 6
    Gemma 3 4B: 6 index, runs locally on Conifer
frontier & open runs locally on Coniferpublished figures · snapshot Jun 2026
  • Slack
  • Gmail
  • GitHub
  • Postgres
  • Messages
  • Calendar
  • Notes
  • Contacts
  • Reminders
  • Files
  • Shell
Marketplace

Someone probably already built what you're looking for.

A market town drawn in ASCII: stalls, carts, and townsfolk around a fountain square. The marketplace is in development.
  • inference providersprove parity, earn routed traffic
  • model labsdebut your model where it runs instantly
  • builderslaunch your agent or app, paid only on sales

download · conifer

Run it on your machine.

Pick your platform. One app, one account, and the free tier is already in the box.

macOS
download
Universal · macOS 11+
Windows
download
x86-64 · Windows 10+
Linux
download
x86-64 · AppImage (.deb also available)