diff --git a/packages/core/src/sync/index.ts b/packages/core/src/sync/index.ts
index a119ccbd74..bf12e4cc32 100644
--- a/packages/core/src/sync/index.ts
+++ b/packages/core/src/sync/index.ts
@@ -13,6 +13,7 @@ import { cloudflareWorkersAi } from "./providers/cloudflare-workers-ai.js";
 import { crossmodel } from "./providers/crossmodel.js";
 import { deepinfra } from "./providers/deepinfra.js";
 import { digitalocean } from "./providers/digitalocean.js";
+import { edenai } from "./providers/edenai.js";
 import { empiriolabs } from "./providers/empiriolabs.js";
 import { google } from "./providers/google.js";
 import { hyper } from "./providers/hyper.js";
@@ -117,6 +118,7 @@ export const providers: {
   crossmodel: SyncProvider<any>;
   deepinfra: SyncProvider<any>;
   digitalocean: SyncProvider<any>;
+  edenai: SyncProvider<any>;
   empiriolabs: SyncProvider<any>;
   google: SyncProvider<any>;
   hyper: SyncProvider<any>;
@@ -143,6 +145,7 @@ export const providers: {
   crossmodel,
   deepinfra,
   digitalocean,
+  edenai,
   empiriolabs,
   google,
   hyper,
@@ -165,6 +168,7 @@ export const providers: {
 export const groups = {
   aggregators: [
     "crossmodel",
+    "edenai",
     "empiriolabs",
     "huggingface",
     "kilo",
diff --git a/packages/core/src/sync/providers/edenai.ts b/packages/core/src/sync/providers/edenai.ts
new file mode 100644
index 0000000000..d17025f467
--- /dev/null
+++ b/packages/core/src/sync/providers/edenai.ts
@@ -0,0 +1,363 @@
+import { existsSync } from "node:fs";
+import path from "node:path";
+import { z } from "zod";
+
+import { describeModel } from "../../describe.js";
+import type {
+  ExistingModel,
+  SyncProvider,
+  SyncedFullModel,
+  SyncedModel,
+} from "../index.js";
+import { factorBaseModel, resolveModelMetadataBaseModel } from "./openrouter.js";
+
+const API_ENDPOINT = "https://api.edenai.run/v3/models";
+const MODELS_DIR = path.join(import.meta.dirname, "..", "..", "..", "..", "..", "models");
+
+// Eden's `owned_by` is a mix of cloud gateways (which proxy someone else's model)
+// and direct model creators. For direct creators we can attempt `<creator>/<model>`
+// resolution; for gateways the model_name already contains the upstream prefix.
+const DIRECT_METADATA_PROVIDER: Record<string, string> = {
+  anthropic: "anthropic",
+  bytedance: "bytedance",
+  cohere: "cohere",
+  deepseek: "deepseek",
+  google: "google",
+  microsoft: "microsoft",
+  minimax: "minimax",
+  mistral: "mistral",
+  moonshot: "moonshotai",
+  openai: "openai",
+  perplexityai: "perplexity",
+  qwen: "alibaba",
+  xai: "xai",
+};
+
+const Pricing = z
+  .object({
+    input_cost_per_token: z.number().optional(),
+    output_cost_per_token: z.number().optional(),
+    cache_read_input_token_cost: z.number().optional(),
+    cache_creation_input_token_cost: z.number().optional(),
+    output_cost_per_reasoning_token: z.number().optional(),
+    input_cost_per_audio_token: z.number().optional(),
+    output_cost_per_image_token: z.number().optional(),
+  })
+  .passthrough();
+
+const Capabilities = z
+  .object({
+    input_modalities: z.array(z.string()).optional(),
+    output_modalities: z.array(z.string()).optional(),
+    supports_reasoning: z.boolean().optional(),
+    supports_tool_choice: z.boolean().optional(),
+    supports_function_calling: z.boolean().optional(),
+    supports_response_schema: z.boolean().optional(),
+    supports_prompt_caching: z.boolean().optional(),
+    supports_vision: z.boolean().optional(),
+    supports_video_input: z.boolean().optional(),
+    supports_audio_input: z.boolean().optional(),
+    supports_pdf_input: z.boolean().optional(),
+  })
+  .passthrough();
+
+export const EdenAIModel = z
+  .object({
+    id: z.string(),
+    created: z.number().optional(),
+    owned_by: z.string(),
+    model_name: z.string(),
+    context_length: z.number().nullable().optional(),
+    description: z.string().nullable().optional(),
+    capabilities: Capabilities,
+    pricing: Pricing.nullable().optional(),
+    alias_of: z.string().nullable().optional(),
+  })
+  .passthrough();
+
+export const EdenAIResponse = z
+  .object({
+    data: z.array(EdenAIModel),
+  })
+  .passthrough();
+
+export type EdenAIModel = z.infer<typeof EdenAIModel>;
+
+export const edenai = {
+  id: "edenai",
+  name: "Eden AI",
+  modelsDir: "providers/edenai/models",
+  preserveBaseModels: false,
+  async fetchModels() {
+    const key = process.env.EDENAI_API_KEY;
+    const response = await fetch(
+      API_ENDPOINT,
+      key ? { headers: { Authorization: `Bearer ${key}` } } : undefined,
+    );
+    if (!response.ok) {
+      throw new Error(
+        `Eden AI request failed: ${response.status} ${response.statusText}`,
+      );
+    }
+    return response.json();
+  },
+  parseModels(raw) {
+    // Aliases duplicate a canonical model under a second ID; skip them so the
+    // catalog stays deduped. The canonical entry still syncs.
+    return EdenAIResponse.parse(raw).data.filter((model) => !model.alias_of);
+  },
+  translateModel(model, context) {
+    if (!isSyncable(model)) return undefined;
+    return {
+      id: model.id,
+      model: buildEdenAIModel(model, context.existing(model.id)),
+    };
+  },
+} satisfies SyncProvider<EdenAIModel>;
+
+function isSyncable(model: EdenAIModel): boolean {
+  // Eden's /v3/models catalog is text-output today. Guard against future entries
+  // (image-gen, TTS) that would need first-class categories in models.dev.
+  const output = model.capabilities.output_modalities;
+  if (output !== undefined && output.length > 0 && !output.includes("text")) {
+    return false;
+  }
+  if (model.context_length == null || model.context_length <= 0) return false;
+  return true;
+}
+
+function baseModelExists(modelID: string): boolean {
+  return existsSync(path.join(MODELS_DIR, `${modelID}.toml`));
+}
+
+function stripRegion(value: string): string {
+  return value.replace(/@[a-z0-9-]+$/i, "");
+}
+
+// Bedrock/gateway names bury the underlying model under provider prefixes and
+// version suffixes: `anthropic.claude-opus-4-6-v1`, `meta.llama3-70b-instruct-v1:0`.
+// Emit progressively-stripped candidates so resolveModelMetadataBaseModel can
+// match the canonical `<lab>/<model>` id.
+function normalizeGatewayName(name: string): string[] {
+  const out = new Set<string>([name]);
+  let s = name.replace(/:\d+$/, "");
+  s = s.replace(/-v\d+$/, "");
+  if (s !== name) out.add(s);
+
+  for (const current of [...out]) {
+    const dotted = current.match(/^([a-z][a-z0-9-]*)\.(.+)$/);
+    if (dotted) {
+      out.add(`${dotted[1]}/${dotted[2]}`);
+      out.add(dotted[2]!);
+    }
+  }
+  return [...out];
+}
+
+function candidateBaseModels(model: EdenAIModel): string[] {
+  const cleanModelName = stripRegion(model.model_name);
+  const cleanID = stripRegion(model.id);
+  const suffix = cleanID.startsWith(`${model.owned_by}/`)
+    ? cleanID.slice(model.owned_by.length + 1)
+    : cleanID;
+
+  const candidates = new Set<string>();
+  const directMeta = DIRECT_METADATA_PROVIDER[model.owned_by];
+  if (directMeta !== undefined) {
+    candidates.add(`${directMeta}/${cleanModelName}`);
+    candidates.add(`${directMeta}/${suffix}`);
+  }
+  candidates.add(cleanModelName);
+  candidates.add(suffix);
+
+  for (const base of [cleanModelName, suffix]) {
+    for (const normalized of normalizeGatewayName(base)) {
+      candidates.add(normalized);
+      if (directMeta !== undefined) {
+        candidates.add(`${directMeta}/${normalized}`);
+      }
+    }
+  }
+  return [...candidates];
+}
+
+function resolveEdenBaseModel(
+  model: EdenAIModel,
+  existingBase: string | undefined,
+): string | undefined {
+  if (existingBase !== undefined && baseModelExists(existingBase)) {
+    return existingBase;
+  }
+  for (const candidate of candidateBaseModels(model)) {
+    if (baseModelExists(candidate)) return candidate;
+    const resolved = resolveModelMetadataBaseModel(candidate);
+    if (resolved !== undefined && baseModelExists(resolved)) return resolved;
+  }
+  return undefined;
+}
+
+type Modality = "text" | "audio" | "image" | "video" | "pdf";
+
+function normalizeModalities(values: string[], fallback: Modality[]): Modality[] {
+  const allowed = new Set<Modality>(["text", "audio", "image", "video", "pdf"]);
+  const mapped = values
+    .map((value) => value.toLowerCase())
+    .map((value) => (value === "file" ? "pdf" : value))
+    .filter((value): value is Modality => allowed.has(value as Modality));
+  return [...new Set(mapped.length > 0 ? mapped : fallback)];
+}
+
+function inputModalitiesFromCapabilities(caps: EdenAIModel["capabilities"]): Modality[] {
+  if (caps.input_modalities !== undefined && caps.input_modalities.length > 0) {
+    return normalizeModalities(caps.input_modalities, ["text"]);
+  }
+  const inferred: Modality[] = ["text"];
+  if (caps.supports_vision) inferred.push("image");
+  if (caps.supports_video_input) inferred.push("video");
+  if (caps.supports_audio_input) inferred.push("audio");
+  if (caps.supports_pdf_input) inferred.push("pdf");
+  return [...new Set(inferred)];
+}
+
+function price(value: number | undefined): number | undefined {
+  if (value === undefined || value === null) return undefined;
+  if (!Number.isFinite(value) || value < 0) return undefined;
+  return Math.round(value * 1_000_000_000_000) / 1_000_000;
+}
+
+function positivePrice(value: number | undefined): number | undefined {
+  const rounded = price(value);
+  return rounded !== undefined && rounded > 0 ? rounded : undefined;
+}
+
+function dateFromTimestamp(timestamp: number | undefined): string | undefined {
+  if (timestamp === undefined) return undefined;
+  const date = new Date(timestamp * 1000);
+  if (Number.isNaN(date.getTime())) return undefined;
+  return date.toISOString().slice(0, 10);
+}
+
+function buildCost(
+  model: EdenAIModel,
+  reasoning: boolean,
+  existing: ExistingModel["cost"] | undefined,
+): SyncedFullModel["cost"] | undefined {
+  const pricing = model.pricing;
+  const input = price(pricing?.input_cost_per_token);
+  const output = price(pricing?.output_cost_per_token);
+  if (input === undefined || output === undefined) return existing;
+  return {
+    input,
+    output,
+    cache_read: positivePrice(pricing?.cache_read_input_token_cost),
+    cache_write: positivePrice(pricing?.cache_creation_input_token_cost),
+    reasoning: reasoning
+      ? positivePrice(pricing?.output_cost_per_reasoning_token) ?? existing?.reasoning
+      : undefined,
+    input_audio: positivePrice(pricing?.input_cost_per_audio_token),
+  };
+}
+
+export function buildEdenAIModel(
+  model: EdenAIModel,
+  existing: ExistingModel | undefined,
+  today = new Date().toISOString().slice(0, 10),
+): SyncedModel {
+  const caps = model.capabilities;
+  const input = inputModalitiesFromCapabilities(caps);
+  const output = normalizeModalities(caps.output_modalities ?? [], ["text"]);
+  const reasoningReported = typeof caps.supports_reasoning === "boolean";
+  const reasoning = caps.supports_reasoning === true;
+  const toolCallReported =
+    typeof caps.supports_function_calling === "boolean" ||
+    typeof caps.supports_tool_choice === "boolean";
+  const toolCall =
+    caps.supports_function_calling === true || caps.supports_tool_choice === true;
+  const structuredOutputReported = typeof caps.supports_response_schema === "boolean";
+  const structuredOutput = caps.supports_response_schema === true;
+  const attachment = input.some((value) => value !== "text");
+  const context = model.context_length ?? 0;
+  // Eden's /v3/models does not return max output tokens. Leave output undefined
+  // for factored files so base_model inheritance provides the authoritative
+  // value; inline files fall back to `context` below to satisfy the schema.
+  const limit = {
+    context,
+    input: existing?.limit?.input,
+    output: existing?.limit?.output,
+  };
+  const inlineReleaseDate =
+    existing?.release_date ?? dateFromTimestamp(model.created) ?? today;
+  const cost = buildCost(model, reasoning, existing?.cost);
+
+  const resolvedBase = resolveEdenBaseModel(model, existing?.base_model);
+  if (resolvedBase !== undefined) {
+    const factored: Partial<SyncedFullModel> = {
+      release_date: existing?.release_date,
+      last_updated: existing?.last_updated ?? today,
+      status: existing?.status,
+      interleaved: existing?.interleaved,
+      cost,
+      limit,
+    };
+    if (existing?.modalities !== undefined) {
+      factored.modalities = existing.modalities;
+      factored.attachment = existing.attachment;
+    }
+    if (existing?.reasoning_options !== undefined) {
+      factored.reasoning_options = existing.reasoning_options;
+    }
+    return factorBaseModel(
+      resolvedBase,
+      factored,
+      limit,
+      existing?.base_model === resolvedBase ? existing.base_model_omit : undefined,
+    );
+  }
+
+  // Inline: schema requires the capability booleans, open_weights, and
+  // limit.output. Fall back to safe defaults when Eden is silent.
+  const inlineReasoning = reasoningReported ? reasoning : existing?.reasoning ?? false;
+  const inlineToolCall = toolCallReported ? toolCall : existing?.tool_call ?? false;
+  const inlineStructuredOutput = structuredOutputReported
+    ? structuredOutput
+    : existing?.structured_output;
+  const openWeights = existing?.open_weights ?? false;
+  const name = existing?.name ?? model.model_name;
+  const description =
+    existing?.description ??
+    (model.description?.trim() ||
+      describeModel({
+        id: model.id,
+        name,
+        family: existing?.family,
+        reasoning: inlineReasoning,
+        tool_call: inlineToolCall,
+        structured_output: inlineStructuredOutput,
+        open_weights: openWeights,
+        limit,
+        modalities: { input, output },
+      }));
+
+  return {
+    name,
+    description,
+    family: existing?.family,
+    attachment,
+    modalities: { input, output },
+    reasoning: inlineReasoning,
+    reasoning_options: inlineReasoning ? existing?.reasoning_options : undefined,
+    tool_call: inlineToolCall,
+    structured_output: inlineStructuredOutput,
+    temperature: existing?.temperature,
+    knowledge: existing?.knowledge,
+    open_weights: openWeights,
+    release_date: inlineReleaseDate,
+    last_updated: existing?.last_updated ?? today,
+    status: existing?.status,
+    interleaved: existing?.interleaved,
+    cost,
+    limit: { ...limit, output: limit.output ?? context },
+    provider: existing?.provider,
+    experimental: existing?.experimental,
+  } satisfies SyncedFullModel;
+}
diff --git a/packages/sdk/src/generated.ts b/packages/sdk/src/generated.ts
index 9847259d1a..56917482ad 100644
--- a/packages/sdk/src/generated.ts
+++ b/packages/sdk/src/generated.ts
@@ -89,8 +89,10 @@ export type ModelFamily =
   | "kimi"
   | "kimi-free"
   | "kimi-k2"
+  | "kimi-k3"
   | "kimi-thinking"
   | "laguna"
+  | "laguna-s"
   | "ling"
   | "ling-flash-free"
   | "liquid"
diff --git a/providers/edenai/logo.svg b/providers/edenai/logo.svg
new file mode 100644
index 0000000000..533f35d718
--- /dev/null
+++ b/providers/edenai/logo.svg
@@ -0,0 +1,3 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 52 64" fill="currentColor">
+  <path fill-rule="evenodd" clip-rule="evenodd" d="M41.1728 21.8429C41.2375 21.1943 41.2707 20.5364 41.2707 19.871C41.2707 18.4667 41.1232 17.0964 40.8426 15.7744C38.9309 6.7654 30.841 0 21.1512 0C10.0395 0 1.03172 8.89653 1.03172 19.871C1.03172 23.3915 1.95872 26.6983 3.58479 29.566C3.85517 30.0429 4.14488 30.5076 4.45286 30.9591C7.94601 28.3218 12.3086 26.7593 17.0244 26.7593C19.8176 26.7593 22.487 27.3075 24.9251 28.3026C27.933 24.2242 32.8146 21.592 38.2761 21.592C39.2632 21.592 40.2313 21.678 41.1728 21.8429ZM40.4619 25.4668C39.7516 25.3468 39.0213 25.2843 38.2762 25.2843C34.2183 25.2843 30.6017 27.1377 28.2573 30.0288C27.4819 30.985 26.8457 32.0548 26.3787 33.2087C25.7987 34.6418 25.4799 36.2046 25.4799 37.8405C25.4799 38.3215 25.5075 38.7962 25.5612 39.2631C32.7086 37.6845 38.4204 32.35 40.4619 25.4668ZM21.899 39.7285C21.8266 39.1095 21.7894 38.4794 21.7894 37.8402C21.7894 35.6007 22.2465 33.4736 23.0717 31.5408C21.322 30.8855 19.429 30.5061 17.4581 30.457C17.3139 30.4534 17.1692 30.4516 17.0242 30.4516C13.1911 30.4516 9.65389 31.6998 6.80839 33.8062C10.4581 37.47 15.536 39.7419 21.1512 39.7419C21.4016 39.7419 21.6509 39.7374 21.899 39.7285Z M23.0717 31.5408C21.322 30.8855 19.429 30.5061 17.4581 30.457C17.3139 30.4534 17.1692 30.4516 17.0242 30.4516C13.1911 30.4516 9.65389 31.6998 6.80839 33.8062C5.81388 34.5424 4.90383 35.3834 4.09521 36.3126C1.54232 39.2461 0 43.0585 0 47.2258C0 50.4128 0.902013 53.3922 2.46821 55.9296C5.45477 60.7683 10.8564 64 17.0242 64C24.2852 64 30.4845 59.521 32.9315 53.214C28.2333 51.6263 24.4695 48.003 22.7703 43.3807C22.3439 42.2208 22.0475 40.9979 21.899 39.7285C21.8266 39.1095 21.7894 38.4794 21.7894 37.8402C21.7894 35.6007 22.2465 33.4736 23.0717 31.5408ZM33.8742 49.6339C33.9889 48.8475 34.0483 48.0435 34.0483 47.2258C34.0483 41.3662 30.9991 36.2083 26.3787 33.2087C25.7987 34.6418 25.4799 36.2046 25.4799 37.8405C25.4799 38.3215 25.5075 38.7962 25.5612 39.2631C25.7053 40.5172 26.0377 41.7154 26.53 42.8298C27.9185 45.9729 30.5789 48.4493 33.8742 49.6339Z M51.0724 37.8405C51.0724 44.775 45.3434 50.3966 38.2762 50.3966C38.1657 50.3966 38.0556 50.3952 37.9458 50.3925C37.7951 50.3888 37.6451 50.3825 37.4958 50.3736C36.2302 50.2989 35.0142 50.0437 33.8742 49.6339C30.5789 48.4493 27.9185 45.9729 26.53 42.8298C26.0377 41.7154 25.7053 40.5172 25.5612 39.2631C25.5075 38.7962 25.4799 38.3215 25.4799 37.8405C25.4799 36.2046 25.7987 34.6418 26.3787 33.2087C26.8457 32.0548 27.4819 30.985 28.2573 30.0288C30.6017 27.1377 34.2183 25.2843 38.2762 25.2843C39.0213 25.2843 39.7516 25.3468 40.4619 25.4668C41.7045 25.6766 42.8858 26.0623 43.9782 26.5967C48.1836 28.6541 51.0724 32.9163 51.0724 37.8405Z"/>
+</svg>
diff --git a/providers/edenai/models/anthropic/claude-haiku-4-5.toml b/providers/edenai/models/anthropic/claude-haiku-4-5.toml
new file mode 100644
index 0000000000..bfcbbe1125
--- /dev/null
+++ b/providers/edenai/models/anthropic/claude-haiku-4-5.toml
@@ -0,0 +1,9 @@
+base_model = "anthropic/claude-haiku-4-5"
+last_updated = "2026-08-04"
+reasoning_options = []
+
+[cost]
+input = 1
+output = 5
+cache_read = 0.1
+cache_write = 1.25
diff --git a/providers/edenai/models/anthropic/claude-opus-4-7.toml b/providers/edenai/models/anthropic/claude-opus-4-7.toml
new file mode 100644
index 0000000000..80b28918e9
--- /dev/null
+++ b/providers/edenai/models/anthropic/claude-opus-4-7.toml
@@ -0,0 +1,9 @@
+base_model = "anthropic/claude-opus-4-7"
+last_updated = "2026-08-04"
+reasoning_options = []
+
+[cost]
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
diff --git a/providers/edenai/models/anthropic/claude-sonnet-4-6.toml b/providers/edenai/models/anthropic/claude-sonnet-4-6.toml
new file mode 100644
index 0000000000..b2c67f9e80
--- /dev/null
+++ b/providers/edenai/models/anthropic/claude-sonnet-4-6.toml
@@ -0,0 +1,9 @@
+base_model = "anthropic/claude-sonnet-4-6"
+last_updated = "2026-08-04"
+reasoning_options = []
+
+[cost]
+input = 3
+output = 15
+cache_read = 0.3
+cache_write = 3.75
diff --git a/providers/edenai/models/cohere/command-a-03-2025.toml b/providers/edenai/models/cohere/command-a-03-2025.toml
new file mode 100644
index 0000000000..4730f55bd1
--- /dev/null
+++ b/providers/edenai/models/cohere/command-a-03-2025.toml
@@ -0,0 +1,9 @@
+base_model = "cohere/command-a-03-2025"
+last_updated = "2026-08-04"
+
+[cost]
+input = 2.5
+output = 10
+
+[limit]
+context = 288_000
diff --git a/providers/edenai/models/cohere/command-r-plus-08-2024.toml b/providers/edenai/models/cohere/command-r-plus-08-2024.toml
new file mode 100644
index 0000000000..beb7057616
--- /dev/null
+++ b/providers/edenai/models/cohere/command-r-plus-08-2024.toml
@@ -0,0 +1,6 @@
+base_model = "cohere/command-r-plus-08-2024"
+last_updated = "2026-08-04"
+
+[cost]
+input = 2.5
+output = 10
diff --git a/providers/edenai/models/deepseek/deepseek-v4-flash.toml b/providers/edenai/models/deepseek/deepseek-v4-flash.toml
new file mode 100644
index 0000000000..18284dd85b
--- /dev/null
+++ b/providers/edenai/models/deepseek/deepseek-v4-flash.toml
@@ -0,0 +1,11 @@
+base_model = "deepseek/deepseek-v4-flash"
+last_updated = "2026-08-04"
+reasoning_options = []
+
+[cost]
+input = 0.14
+output = 0.28
+cache_read = 0.0028
+
+[limit]
+context = 1_048_576
diff --git a/providers/edenai/models/deepseek/deepseek-v4-pro.toml b/providers/edenai/models/deepseek/deepseek-v4-pro.toml
new file mode 100644
index 0000000000..edb801b5cb
--- /dev/null
+++ b/providers/edenai/models/deepseek/deepseek-v4-pro.toml
@@ -0,0 +1,11 @@
+base_model = "deepseek/deepseek-v4-pro"
+last_updated = "2026-08-04"
+reasoning_options = []
+
+[cost]
+input = 0.435
+output = 0.87
+cache_read = 0.003625
+
+[limit]
+context = 1_048_576
diff --git a/providers/edenai/models/google/gemini-2.5-flash.toml b/providers/edenai/models/google/gemini-2.5-flash.toml
new file mode 100644
index 0000000000..94657a95d8
--- /dev/null
+++ b/providers/edenai/models/google/gemini-2.5-flash.toml
@@ -0,0 +1,11 @@
+base_model = "google/gemini-2.5-flash"
+last_updated = "2026-08-04"
+reasoning_options = []
+
+[cost]
+input = 0.3
+output = 2.5
+reasoning = 2.5
+cache_read = 0.03
+cache_write = 0.083333
+input_audio = 1
diff --git a/providers/edenai/models/google/gemini-2.5-pro.toml b/providers/edenai/models/google/gemini-2.5-pro.toml
new file mode 100644
index 0000000000..9d53a95ee4
--- /dev/null
+++ b/providers/edenai/models/google/gemini-2.5-pro.toml
@@ -0,0 +1,11 @@
+base_model = "google/gemini-2.5-pro"
+last_updated = "2026-08-04"
+reasoning_options = []
+
+[cost]
+input = 1.25
+output = 10
+reasoning = 10
+cache_read = 0.125
+cache_write = 0.375
+input_audio = 1.25
diff --git a/providers/edenai/models/minimax/MiniMax-M3.toml b/providers/edenai/models/minimax/MiniMax-M3.toml
new file mode 100644
index 0000000000..bca3538de0
--- /dev/null
+++ b/providers/edenai/models/minimax/MiniMax-M3.toml
@@ -0,0 +1,11 @@
+base_model = "minimax/MiniMax-M3"
+last_updated = "2026-08-04"
+reasoning_options = []
+
+[cost]
+input = 0.3
+output = 1.2
+cache_read = 0.06
+
+[limit]
+context = 524_288
diff --git a/providers/edenai/models/mistral/codestral-latest.toml b/providers/edenai/models/mistral/codestral-latest.toml
new file mode 100644
index 0000000000..daddbf8ac5
--- /dev/null
+++ b/providers/edenai/models/mistral/codestral-latest.toml
@@ -0,0 +1,6 @@
+base_model = "mistral/codestral-latest"
+last_updated = "2026-08-04"
+
+[cost]
+input = 1
+output = 3
diff --git a/providers/edenai/models/mistral/mistral-large-latest.toml b/providers/edenai/models/mistral/mistral-large-latest.toml
new file mode 100644
index 0000000000..01c90d63ff
--- /dev/null
+++ b/providers/edenai/models/mistral/mistral-large-latest.toml
@@ -0,0 +1,7 @@
+base_model = "mistral/mistral-large-latest"
+last_updated = "2026-08-04"
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.2
diff --git a/providers/edenai/models/moonshot/kimi-k3.toml b/providers/edenai/models/moonshot/kimi-k3.toml
new file mode 100644
index 0000000000..1210c131bc
--- /dev/null
+++ b/providers/edenai/models/moonshot/kimi-k3.toml
@@ -0,0 +1,8 @@
+base_model = "moonshotai/kimi-k3"
+last_updated = "2026-08-04"
+reasoning_options = []
+
+[cost]
+input = 3
+output = 15
+cache_read = 0.3
diff --git a/providers/edenai/models/openai/gpt-4.1.toml b/providers/edenai/models/openai/gpt-4.1.toml
new file mode 100644
index 0000000000..5494ade32b
--- /dev/null
+++ b/providers/edenai/models/openai/gpt-4.1.toml
@@ -0,0 +1,7 @@
+base_model = "openai/gpt-4.1"
+last_updated = "2026-08-04"
+
+[cost]
+input = 2
+output = 8
+cache_read = 0.5
diff --git a/providers/edenai/models/openai/gpt-4o-mini.toml b/providers/edenai/models/openai/gpt-4o-mini.toml
new file mode 100644
index 0000000000..7fd090546e
--- /dev/null
+++ b/providers/edenai/models/openai/gpt-4o-mini.toml
@@ -0,0 +1,7 @@
+base_model = "openai/gpt-4o-mini"
+last_updated = "2026-08-04"
+
+[cost]
+input = 0.15
+output = 0.6
+cache_read = 0.075
diff --git a/providers/edenai/models/openai/gpt-5.6-terra.toml b/providers/edenai/models/openai/gpt-5.6-terra.toml
new file mode 100644
index 0000000000..ba95166b91
--- /dev/null
+++ b/providers/edenai/models/openai/gpt-5.6-terra.toml
@@ -0,0 +1,9 @@
+base_model = "openai/gpt-5.6-terra"
+last_updated = "2026-08-04"
+reasoning_options = []
+
+[cost]
+input = 2
+output = 12
+cache_read = 0.2
+cache_write = 2.5
diff --git a/providers/edenai/models/openai/o3.toml b/providers/edenai/models/openai/o3.toml
new file mode 100644
index 0000000000..1e2474be33
--- /dev/null
+++ b/providers/edenai/models/openai/o3.toml
@@ -0,0 +1,8 @@
+base_model = "openai/o3"
+last_updated = "2026-08-04"
+reasoning_options = []
+
+[cost]
+input = 2
+output = 8
+cache_read = 0.5
diff --git a/providers/edenai/models/xai/grok-4.5.toml b/providers/edenai/models/xai/grok-4.5.toml
new file mode 100644
index 0000000000..f6382a6cab
--- /dev/null
+++ b/providers/edenai/models/xai/grok-4.5.toml
@@ -0,0 +1,8 @@
+base_model = "xai/grok-4.5"
+last_updated = "2026-08-04"
+reasoning_options = []
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.3
diff --git a/providers/edenai/provider.toml b/providers/edenai/provider.toml
new file mode 100644
index 0000000000..e07e014bfe
--- /dev/null
+++ b/providers/edenai/provider.toml
@@ -0,0 +1,5 @@
+name = "Eden AI"
+env = ["EDENAI_API_KEY"]
+npm = "@ai-sdk/openai-compatible"
+api = "https://api.edenai.run/v3"
+doc = "https://www.edenai.co/docs"
\ No newline at end of file